using truck fleet data in combination with other …...using truck fleet data in combination with...

Using Truck Fleet Data in Combination with Other Data Sources for

Freight Modeling and Planning

BDK84-977-20

Final Report

Prepared for:

Florida Department of Transportation

Mr. Frank Tabatabaee, Project Manager

Prepared by:

University of South Florida

Dr. Abdul R. Pinjari

Mr. Akbar Bakhshi Zanjani

Mr. Aayush Thakur

Ms. Anissa Nur Irmania

Mr. Mohammadreza Kamali

American Transportation Research Institute

Mr. Jeffrey Short

Mr. Dave Pierce

Ms. Lisa Park

July 2014

ii

DISCLAIMER

The opinions, findings, and conclusions expressed in this publication are those of the authors and

not necessarily those of the State of Florida Department of Transportation.

iii

METRIC CONVERSION

SYMBOL WHEN YOU KNOW MULTIPLY BY TO FIND SYMBOL

LENGTH

in inches 25.4 millimeters mm

ft feet 0.305 meters m

yd yards 0.914 meters m

mi miles 1.61 kilometers km

iv

TECHNICAL REPORT DOCUMENTATION

1. Report No.

2. Government Accession No.

3. Recipient's Catalog No.

4. Title and Subtitle

Using Truck Fleet Data in Combination with Other Data Sources for

Freight Modeling and Planning

5. Report Date

July 2014

6. Performing Organization Code

USF

7. Author(s)

Abdul R. Pinjari, Akbar Bakhshi Zanjani, Aayush Thakur, Anissa

Irmania, Mohammadreza Kamali, Jeffrey Short, Dave Pierce, Lisa Park

8. Performing Organization Report No.

9. Performing Organization Name and Address

College of Engineering

Department of Civil and Environmental Engineering

University of South Florida

4202 E. Fowler Avenue, ENB118, Tampa FL 33620

10. Work Unit No. (TRAIS)

11. Contract or Grant No.

BDK84-977-20

12. Sponsoring Agency Name and Address

Florida Department of Transportation

Research Center

605 Suwannee Street, MS 30

Tallahassee, FL 32399-0450

13. Type of Report and Period Covered

Final Report

5/3/2011– 6/3/2014

14. Sponsoring Agency Code

15. Supplementary Notes

16. Abstract

This project investigated the use of large streams of truck GPS data available from the American Transportation

Research Institute (ATRI) for the following statewide freight modeling and planning applications in Florida:

(1) Average truck speed data were developed for each (and every) mile of Florida’s Strategic Intermodal System

(SIS) highway network for different time periods in the day.

(2) Algorithms were developed to convert raw truck GPS data into a database of truck trips. The algorithms were

applied to convert four months of raw data, comprising 145 million GPS records, into more than 1.2 million

truck trips traveling within, into, and out of Florida.

(3) The truck trip database developed from ATRI’s truck GPS data was used to analyze truck travel

characteristics, including trip duration, trip length, trip speed, and time-of-day profiles, for different regions

in the state. Further, distributions of origin-destination (OD) truck travel times were derived for more than

1,200 OD pairs in the Florida Statewide Model (FLSWM).

(4) ATRI’s truck GPS data were evaluated for their coverage of truck traffic flows in Florida. At an aggregate

level, the data were found to capture 10 percent of heavy truck volumes observed in Florida.

(5) The data were used to develop OD tables of statewide freight truck flows within, into, and out of Florida. To

do so, the truck trip database developed from ATRI’s truck GPS data was combined with observed truck

traffic volumes at different locations in Florida and other states using OD matrix estimation procedures. The

OD tables were developed for the spatial resolution of traffic analysis zones used in FLSWM.

(6) Preliminary explorations were conducted with ATRI’s truck GPS data for analyzing truck travel routes

between different locations and for analyzing truck flows out of seaports.

17. Key Word

Freight data, truck GPS data, freight performance

measures, OD flow table, statewide freight travel

demand models

18. Distribution Statement

19. Security Classif. (of this report)

Unclassified

20. Security Classif. (of this page)

Unclassified

21. No. of Pages

238

22. Price

v

ACKNOWLEDGMENTS

The authors would like to express their since appreciation to the Florida Department of

Transportation (FDOT) managers on this project, Mr. Vidya Mysore and Mr. Frank Tabatabaee,

for their full support, constant guidance, and valuable feedback during the project. Mr. Mysore,

former manager of systems traffic modeling at FDOT, spearheaded the project and provided

valuable input and feedback. Mr. Tabatabaee served as the project manager in the later stages of

the project and played a key role in helping us complete the project. Thanks are due to Mr.

Thomas Hill, current manager of the FDOT Systems Planning Office, for his support to the

project. Mr. Frank Tabatabaee and Mr. Vladimir Majano provided the research team access to

various FDOT datasets required for the project. Thanks to Mr. Daniel Lamb, Ms. Denise

Bunnewith, and Mr. Larry Foutz for their input as project panel members. The authors would

like to thank the Florida Model Task Force (MTF) members for their input and discussion during

the project presentations made at several MTF meetings.

Special thanks are due to Mr. Dan Murray of the American Transportation Research

Institute (ATRI) for his input and collaboration on the project. Thanks to Dr. Vince Bernardin

from Resource Systems Group and Dr. Arun Kuppam from Cambridge Systematics for sharing

their experience in working with ATRI’s truck GPS data. Thanks to Mr. Vipul Modi from

Citilabs for his assistance with the use of Cube software for the origin-destination matrix

estimation procedure. Thanks to Dr. Ramachandran Balakrishna and Dr. Howard Slavin from

Caliper Corporation for helpful insights on the theoretical and practical aspects of origin-

destination matrix estimation. Thanks to Mr. Krishnan Viswanathan from CDM Smith and Mr.

Colin Smith and Ms. Maren Outwater from Resource Systems Group for their input at different

stages of the project.

The authors would like to thank Mr. Vaibhav Rastogi and Mr. Anirban Ghosh, who

implemented several algorithms developed for the project in Java programming platform. Mr.

Bertho Augustin and Mr. Jonathan Koons provided valuable assistance with the preparation of

different datasets and validation of outputs from the procedures developed in the project. Ms.

Cherise Burton and Mr. Akshay Kancharla assisted in some of the validation activities. Finally,

the authors thank Ms. Patricia Ball and Ms. Vicki Morrison for their editorial review of this

report.

vi

EXECUTIVE SUMMARY

An accelerated growth in the volume of freight shipped on American highways has led to a

significant increase in truck traffic, influencing traffic operations, safety, and the state of repair

of highway infrastructure. Traffic congestion, in turn, has impeded the speed and reliability of

freight movement on the highway system. As freight movement continues to grow within and

between urban areas, appropriate planning and decision making processes are necessary to

mitigate the above-mentioned impacts. However, a main challenge in establishing these

processes is the lack of adequate data on freight movements such as detailed origin-destination

(OD) data, truck travel times, freight tonnage distribution by OD pairs, transported commodity

by OD pairs, and details about truck trip stops and paths. As traditional data sources on freight

movement are either inadequate or no longer available, new sources of data must be investigated.

A recently-available source of data on nationwide freight flows is based on a joint

venture by the American Transportation Research Institute (ATRI) and the Federal Highway

Administration (FHWA) to develop and test a national system for monitoring freight

performance measures (FPM) on key corridors in the nation. These data are obtained from

trucking companies that use GPS-based technologies to remotely monitor their trucks. ATRI’s

truck Geographical Position System (GPS) database contains GPS traces of a large number of

trucks as they traveled through the national highway system. This provides unprecedented

amounts of data on freight truck movements throughout the nation (and Florida). Such truck GPS

data potentially can be used to support planning, operation, and management processes

associated with freight movements. Further, the data can be put to better use when used in

conjunction with other freight data obtained from other sources.

The overarching goal of this project is to investigate the use of ATRI’s truck GPS data

for statewide freight performance measurement, statewide freight truck flow analysis, and other

freight planning and modeling applications in the state. The specific objectives are to:

1) Derive freight performance measures for Florida’s Strategic Intermodal System (SIS)

highways,

2) Develop algorithms to convert large streams of ATRI’s truck GPS data into a more

useable truck trip format,

3) Analyze truck trip characteristics in Florida using ATRI’s truck GPS data,

4) Assess ATRI’s truck GPS data in terms of its coverage of truck traffic flows in

Florida,

5) Develop OD tables of statewide freight truck flows within, into, and out of Florida for

different geographic resolutions, including the Florida Statewide Model (FLSWM)

traffic analysis zones (TAZs), and

6) Explore the use of ATRI’s GPS data for other applications of interest to Florida,

including the analysis of truck flows out of two seaports, the re-routing patterns of

trucks after a major highway incident, and the routing patterns of trucks traveling

between Jacksonville and Ocala.

vii

Project Outcomes and Findings

The outcomes and findings from the project are discussed next.

Freight Performance Measures on Florida’s SIS Highway Network

The project resulted in the development of average truck speed data for each (and every) mile of

the Strategic Intermodal System (SIS) highway network for different time periods in the day—

AM peak, PM peak, mid-day, and off-peak—using three months of ATRI’s truck GPS data in

the year 2010. In doing so, it was found that the existing shape files of the SIS network available

from FDOT either were not accurate enough or they lacked the details (for example, separate

links by direction for divided highways) to derive performance measures using geospatial data.

Therefore, a highly accurate network was developed to represent highways on the SIS network.

The SIS highway network shape file and the data on average truck speeds by time-of-day

were submitted in a GIS shape file that can be used in an ArcGIS environment to identify the

major freight bottlenecks on Florida’s SIS highway network. In addition to the development of

average speed measures, the project developed example applications of ATRI’s truck GPS data

for measuring truck speed reliability and analyzing highway freight bottlenecks.

Algorithms to Convert ATRI’s Raw GPS Data Streams into a Database of Truck Trips

The raw GPS data streams from ATRI need to be converted into a truck trip format to realize the

full potential of the data for freight planning applications. The project resulted in algorithms to

convert the raw GPS data into a database of truck trips. The results from the algorithms were

subjected to different validations to confirm that the algorithms can be used to extract accurate

trip information from raw GPS data provided by ATRI.

These algorithms were then applied to four months of raw GPS data from ATRI,

comprising a total of 145 million raw GPS data records, to develop a large database of truck trips

traveling within, into, and out of the state. The resulting database comprised more than 1.2

million truck trips traveling within, into, and out of the state. This database of truck trips can be

used for a variety of purposes, including the development of truck travel characteristics and OD

truck flow patterns for different geographical regions in Florida. The database can be used to

calibrate and validate the next-generation statewide freight travel demand model being

developed by FDOT. In future work, this database potentially can be used to develop data on

truck trip-chaining and logistics patterns in the state.

Analysis of Truck Travel Characteristics in Florida

The truck trip database developed from four months of ATRI’s truck GPS data was used to

analyze a variety of truck travel characteristics in the state of Florida. The truck travel

characteristics analyzed include trip duration, trip length, trip speed, time-of-day profiles, and

OD flows. Each of these characteristics was derived at a statewide level as well as for different

regions in the state—Jacksonville, Tampa Bay, Orlando, Miami, and rest of Florida—defined

based on the freight analysis framework (FAF) zoning system.

In addition, the truck trips were used in conjunction with the GPS data to derive

distributions of OD travel distances, travel times, and travel speeds between more than 1,200

TAZ-to-TAZ OD pairs in the FLSWM. Comparing the minimum truck travel times measured

viii

using GPS data for the 1,200 OD pairs with free flow travel times used as inputs to FLSWM

indicated that the FLSWM travel times are systematically underestimated when compared to the

truck travel times measured from ATRI data. A similar comparison with the travel times

extracted from Google Maps suggested that the Google Maps travel times also underestimate

(albeit not as much the FLSWM travel times) truck travel times between origins and

destinations. This is most likely because the travel times used as inputs for the FLSWM and

those reported by Google Maps are predominantly geared toward passenger cars that tend to

have higher travel speeds and better acceleration characteristics. ATRI’s truck GPS data, on the

other hand, provide an opportunity to accurately measure travel times exclusively for trucks (and

for different times of the day).

In addition to the measurement of OD truck travel distances, travel times, and speeds, the

project team performed an exploratory analysis of truck travel routes for more than 1,600 trips

between 10 OD pairs in FLSWM. A preliminary exploratory analysis suggested that a majority

of trips between any OD pair tend to travel by largely similar routes (i.e., the variability in route

choice is not high for the 10 OD pairs examined in this study). Specifically, considerable overlap

was observed among the routes across a large number of trips between an OD pair. This

observation has interesting implications for future research on understanding and modeling truck

route choice. While this study did not delve further into understanding the route choice patterns

of long-haul trucks, this is an important area for future research using the truck GPS data from

ATRI.

Assessment of ATRI’s Truck GPS Data and Its Coverage of Truck Traffic in Florida

This project resulted in a better understanding of ATRI’s truck GPS data in terms of its coverage

of truck traffic in the state of Florida. This includes deriving insights on (a) the types of trucks

(e.g., heavy trucks and medium trucks) present in the data, (b) the geographical coverage of the

data in Florida, and (c) the proportion of the truck traffic flows in the state covered by the data.

Based on discussions with ATRI and anecdotal evidence, it is known that the major

sources of ATRI data are freight shipping companies that own large trucking fleets, which

typically comprise tractor-trailer combinations (or Federal Highway Administration [FHWA]

vehicle type classes 8–13). However, a close observation of the data, from following the trucks

on Google Earth and examining travel characteristics of individual trucks, suggests that the data

include a small proportion of trucks that are likely to be smaller trucks that do not necessarily

haul freight over long distances. The project used simple rules to divide the data into two

categories: (1) long-haul trucks or heavy trucks (considered to be FHWA vehicle classes 8–13),

and (2) short-haul trucks or medium trucks. Specifically, trucks that did not make at least one trip

of 100-mile length in a two-week period and those that made more than 5 trips per day were

considered short-haul or medium trucks. Out of a total of 169,714 unique truck IDs in the data,

about 4.6 percent were labeled as short-haul trucks (or medium trucks) and separated from the

remaining long-haul trucks (or heavy trucks). In future work, it will be useful to derive better

definitions of heavy trucks and medium trucks. Whereas heavy trucks are of primary interest to

FLSWM for updating and validating its freight truck model components (assuming these are the

long-haul freight carrying trucks), medium trucks are also of potential use for updating the non-

freight truck model components in FLSWM. Further, extracting sufficient data on medium trucks

ix

potentially can help understand truck movement within urban regions as well, because a

considerable proportion of truck traffic in urban areas tends to comprise medium trucks.

ATRI’s truck GPS data represent a sample of truck flows within, coming into, and going

out of Florida. This sample is not a census of all trucks traveling in the state. Also, it is unknown

what proportion of heavy truck flows in the state is represented by this data sample. To address

this question, truck traffic flows implied by one-week of ATRI’s truck GPS data were compared

with truck counts data from more than 200 Telemetered Traffic Monitoring Sites (TTMS) in the

state. The results from this analysis suggest that, at an aggregate level, the ATRI data provide

10.1 percent coverage of heavy truck flows observed in Florida. When the coverage was

examined separately for different highway facilities (based on functional classification), the

results suggest that the data provide a representative coverage of truck flows through different

types of highway facilities in the state.

The coverage of ATRI data was examined for different geographical regions in Florida

by examining the spatial distribution of the number of truck trips generated at TAZ-level and at

county-level geography. In addition, the percentage of heavy truck traffic covered by ATRI data

at different locations was examined. All these examinations suggest potential geographical

differences in the extent to which ATRI data represent heavy truck traffic volumes at different

locations in the state. For instance, truck trips generated from Polk County were much higher

than those generated from Hillsborough and Miami-Dade counties. Further, the percentage of

heavy truck traffic covered by ATRI data in the southern part of Florida (within Miami) and the

southern stretch of I-75 is slightly lower compared to the coverage in the northern and central

Florida regions. Such geographical differences (or spatial biases) potentially can be adjusted by

combining ATRI’s truck GPS data with observed data on truck traffic flows at different locations

in the state (from FDOT’s TTMS traffic counting program).

OD Tables of Statewide Truck Flows

An important outcome of the project was to use ATRI’s truck GPS data in combination with

other available data to derive OD tables of freight truck flows within, into, and out of the state of

Florida. The OD flow tables were derived at the following levels of geographic resolution:

a) TAZs of the FLSWM, where Florida and the rest of the country are divided into about

6,000 TAZs,

b) County-level resolution, where Florida is represented at a county-level resolution and

the rest of the country is represented at a state-level resolution, and

c) State-level resolution, where Florida and the rest of the country are represented at a

state-level resolution.

As part of this task, first, the truck trip database developed from four months of ATRI’s

GPS data was converted into OD tables at the TAZ-level spatial resolution used in the FLSWM.

Such an OD table derived only from the ATRI data, however, is not necessarily representative of

the freight truck flows in the state. This is because the ATRI data does not comprise the census

of trucks in the state. Besides, it is not necessarily a random sample and is likely to have spatial

biases in its representation of truck flows in the state. To address these issues, the OD tables

derived from the ATRI data were combined with observed truck traffic volumes at different

x

locations in the state (and outside the state) to derive a more robust OD table that is

representative of the freight truck flows within, into, and out of the state. To achieve this, a

mathematical procedure called the Origin-Destination Matrix Estimation (ODME) method was

employed to combine the OD flow table generated from the ATRI data with observed truck

traffic volume information at different locations within and outside Florida. The OD flow table

estimated from the ODME procedure is likely to better represent the heavy truck traffic volumes

in the state, as it uses the observed truck traffic volumes to weigh the ATRI data-derived truck

OD flow tables.

Explorations of the Use of ATRI’s Truck GPS Data for Other Applications

In addition to the above, this project conducted preliminary explorations of the use of ATRI’s

truck GPS data for the following applications:

a) Analysis of truck flows out of two ports in Florida—Port Blount Island in

Jacksonville and Port Everglades in Fort Lauderdale,

b) Analysis of routing patterns of trucks that used the US 301 roadway to travel between

I-95 around Jacksonville and I-75 around Ocala, and

c) Analysis of changes in truck routing patterns during the closure of a stretch of I-75

near Ocala due to a major multi-vehicle crash in January 2012.

Note that these applications were only preliminary explorations conducted as proofs of

concept. Future work can expand on these explorations to conduct full-scale applications.

xi

TABLE OF CONTENTS

DISCLAIMER ................................................................................................................................ ii

METRIC CONVERSION .............................................................................................................. iii

TECHNICAL REPORT DOCUMENTATION ............................................................................ iv

ACKNOWLEDGMENTS .............................................................................................................. v

EXECUTIVE SUMMARY ........................................................................................................... vi

Project Outcomes and Findings ........................................................................................ vii

Freight Performance Measures on Florida’s SIS Highway Network ................... vii

Algorithms to Convert ATRI’s Raw GPS Data Streams into a Database of

Truck Trips...................................................................................................... vii

Analysis of Truck Travel Characteristics in Florida ............................................. vii

Assessment of ATRI’s Truck GPS Data and Its Coverage of Truck Traffic

in Florida ........................................................................................................ viii

OD Tables of Statewide Truck Flows .................................................................... ix

Explorations of the Use of ATRI’s Truck GPS Data for Other Applications ........ x

LIST OF FIGURES ...................................................................................................................... xv

LIST OF TABLES ...................................................................................................................... xxv

CHAPTER 1: INTRODUCTION ................................................................................................... 1

1.1 Background ................................................................................................................... 1

1.2 Project Objectives ......................................................................................................... 2

1.2.1 Objective 1: Derive Freight Performance Measures for Florida’s

Freight-intensive Highways .............................................................................. 2

1.2.2 Objective 2: Develop Methods to Convert ATRI’s Raw GPS Data

Streams into Truck Trips ................................................................................. 2

1.2.3 Objective 3: Analyze Truck Travel Characteristics in Florida ...................... 3

1.2.4 Objective 4: Assess ATRI’s Truck GPS Data and Its Coverage of Truck

Traffic in Florida .............................................................................................. 3

1.2.5 Objective 5: Derive Statewide Truck Trip Flow OD Tables for the

Traffic Analysis Zone (TAZ)-level Spatial Resolution in the Florida

Statewide Model (FLSWM) ............................................................................. 3

1.2.6 Objective 6: Explore the Use of ATRI’s Truck GPS Data for Other

Applications of Interest to Florida ................................................................... 4

1.3 Structure of the Report .................................................................................................. 4

CHAPTER 2: OVERVIEW OF ATRI DATA AND ITS APPLICATIONS FOR FREIGHT

PERFORMANCE MEASUREMENT...................................................................................... 5

2.1 Introduction ................................................................................................................... 5

2.2 Background on ATRI's Truck GPS Data ...................................................................... 5

2.3 Applications of ATRI’s Truck GPS Data for Freight Performance Measurement ....... 9

xii

2.3.1 Freight Performance Measures: Average Truck Speeds on SIS Highways ... 9

2.3.2 Freight Performance Measures: Truck Speed Reliability Measurements .... 13

2.3.3 Identification and Analysis of Freight Bottlenecks ..................................... 14

CHAPTER 3: CONVERSION OF ATRI TRUCK GPS DATA TO TRUCK TRIPS ................. 17

3.1 Introduction ................................................................................................................. 17

3.2 ATRI’s Truck GPS Data ............................................................................................. 17

3.3 Algorithm Description ................................................................................................ 18

3.3.1 Clean and Sort Data ..................................................................................... 19

3.3.2 Identify Truck Stops (i.e., Truck Trip-ends) to Generate Truck Trips ........ 19

3.3.2.1 Determining Truck Stops and Moving Instances .......................... 21

3.3.2.2 Separating Intermediate Stops from Trip Destinations ................. 22

3.3.2.3 Dealing with Insignificant Trips ................................................... 23

3.3.2.4 Quality Control Checks in the Algorithm ..................................... 23

3.3.3 Eliminate Trip Ends in Rest Areas............................................................... 24

3.3.4 Find Circular Trips and Split Them into Shorter Valid Trips ...................... 25

3.4 Results ......................................................................................................................... 27

CHAPTER 4: CHARACTERISTICS OF TRUCK TRIPS DERIVED FROM ATRI DATA .... 30

4.1 Introduction ................................................................................................................. 30

4.2 Trip Length, Trip Time, Trip Speed, and Time-of-Day Distributions........................ 30

4.3 Origin-Destination Truck Travel Time and Route Measurements ............................. 42

4.3.1 Factors Influencing Travel Distance and Travel Time Measurements

from ATRI’s GPS Data ................................................................................... 42

4.3.2 Truck Travel Distance, Travel Time, and Travel Route Measurements

for 1,237 OD pairs .......................................................................................... 44

CHAPTER 5: ASSESSMENT OF THE COVERAGE OF TRUCK FLOWS IN FLORIDA

BY THE ATRI DATA ............................................................................................................ 50

5.1 Introduction ................................................................................................................. 50

5.2 Truck Types in ATRI Data ......................................................................................... 50

5.3 What Proportion of Heavy Truck Traffic Flows in Florida Is Captured in ATRI’s

Truck GPS Data? ........................................................................................................ 51

5.4 Geographical Coverage of ATRI’s Data in Florida .................................................... 55

CHAPTER 6: ESTIMATION OF STATEWIDE ORIGIN-DESTINATION TRUCK

FLOWS ................................................................................................................................... 59

6.1 Introduction ................................................................................................................. 59

6.2 Origin-Destination Matrix Estimation (ODME) ......................................................... 59

6.3 Inputs and Assumptions for the ODME Procedure .................................................... 62

6.3.1 The Seed Matrix ........................................................................................... 62

6.3.1.1 Geographical Coverage of the Seed Matrix .................................. 63

6.3.1.2 Zero Cells in the Seed Matrix ....................................................... 67

6.3.2 Truck Traffic Counts.................................................................................... 68

6.3.2.1 Truck Traffic Counts in Florida .................................................... 68

6.3.2.2 Truck Traffic Counts outside Florida ........................................... 70

xiii

6.3.3 Network........................................................................................................ 71

6.4 Results from the ODME Procedure ............................................................................ 71

6.4.1 Evaluation of Different Assumptions .......................................................... 71

6.4.2 ODME Results for One Set of Assumptions ............................................... 74

6.5 Scope for Improvements to ODME Results ............................................................... 81

CHAPTER 7: SUMMARY AND CONCLUSIONS .................................................................... 84

7.1 Summary ..................................................................................................................... 84

7.2 Project Outcomes and Findings .................................................................................. 85

7.2.1 Freight Performance Measures on Florida’s SIS Highway Network .......... 85

7.2.2 Algorithms to Convert ATRI’s Raw GPS Data Streams into a Database

of Truck Trips ................................................................................................. 85

7.2.3 Analysis of Truck Travel Characteristics in Florida .................................... 86

7.2.4 Assessment of ATRI’s Truck GPS Data and Its Coverage of Truck

Traffic in Florida ............................................................................................. 86

7.2.5 Origin-Destination (OD) Tables of Statewide Truck Flows ........................ 88

7.2.6 Explorations of the Use of ATRI’s Truck GPS Data for Other

Applications .................................................................................................... 88

7.3 Future Work ................................................................................................................ 89

7.3.1 Explore the Use of ATRI’s Truck GPS Data for Understanding Urban

Freight Movements and Statewide Non-freight Truck Flows ........................ 89

7.3.2 Estimation of OD Truck Travel Times for FLSWM ................................... 89

7.3.3 Analysis of Truck Route Choice .................................................................. 89

7.3.4 Improvements to ODME.............................................................................. 90

7.3.5 Development of Truck Trip Chaining and Logistics Data ........................... 90

REFERENCES ............................................................................................................................. 91

APPENDIX A: EXPLORATORY ANALYSES OF USING ATRI DATA FOR FREIGHT

PLANNING APPLICATIONS IN FLORIDA ....................................................................... 93

A.1 I-75 Incident Analysis ................................................................................................ 93

A.2 Trucks Flows from Ports ............................................................................................ 95

A.2.1 Blount Island Analysis ................................................................................ 95

A.2.2 Port Everglades Analysis ............................................................................ 97

A.3 I-75 Ocala Analysis .................................................................................................... 99

APPENDIX B: DISTRIBUTIONS OF TRIP LENGTH, TRIP DURATION, TRIP

SPEED, AND TRIP TIME-OF-DAY FOR DIFFERENT SEGMENTS OF TRUCK

TRIPS WITHIN, TO, AND FROM FLORIDA ................................................................... 101

Characteristics of Truck Trips within the Tampa FAF Zone .......................................... 102

Characteristics of Truck Trips within the Orlando FAF Zone ........................................ 108

Characteristics of Truck Trips within the Jacksonville FAF Zone ................................. 114

Characteristics of Truck Trips within the Miami FAF Zone .......................................... 120

Characteristics of Truck Trips Starting and Ending in the Remainder of Florida .......... 126

Characteristics of Truck Trips Starting and Ending in Florida (I-I Trips for Florida) ... 132

xiv

Characteristics of Truck Trips Starting in Florida and Ending outside Florida

(I-E Trips) ................................................................................................................. 136

Characteristics of Truck Trips Starting outside Florida and Ending in Florida

(E-I Trips) ................................................................................................................. 141

APPENDIX C: TRAVEL ROUTES AND TRAVEL TIME MEASUREMENTS FOR

10 OD PAIRS ....................................................................................................................... 146

APPENDIX D: SEASONALITY ANALYSIS .......................................................................... 171

D.1 Single-Unit Trucks ................................................................................................... 172

D.2 Tractor-Trailer Trucks.............................................................................................. 192

xv

LIST OF FIGURES

Figure 2.1 FHWA Vehicle Classifications ..................................................................................... 7

Figure 2.2 Sample of Truck GPS Record Positions in Florida ...................................................... 8

Figure 2.3 Truck GPS Record Positions throughout North America for Same Sample

of Trucks in Florida from Figure 2.2 ............................................................................ 9

Figure 2.4 Average Truck Speeds for AM Peak Time Period on Florida’s SIS Highways ......... 11

Figure 2.5 Average Truck Speeds for Different Time Periods in Day for One-Mile

Segment in Miami Region .......................................................................................... 12

Figure 2.6 Average Speeds by Hour for Miles 24–48 on I-95 in Fort Lauderdale, Florida ......... 13

Figure 2.7 Highway Corridors Selected for Truck Speed Reliability Measurements .................. 14

Figure 2.8 Fusion of ATRI’s Freight Performance Measures FAF Data on Truck Volumes

to Identify Important Bottlenecks for Freight Movement ........................................... 15

Figure 2.9 Bottleneck Analysis for I-4 at I-275 in Tampa ........................................................... 16

Figure 3.1 Algorithm for Identifying Truck Trip Ends from Raw GPS Data .............................. 20

Figure 3.2 Example of Circular Trip Extracted from Algorithm with 30-Minute

Dwell-time Buffer ....................................................................................................... 26

Figure 4.1 Trip Length Distribution of All Trips Derived from Four Months of ATRI Data ..... 31

Figure 4.2 Trip Time Distribution of All Trips Derived from Four Months of ATRI Data ........ 31

Figure 4.3 Trip Speed Distribution of All Trips Derived from Four Months of ATRI Data ....... 32

Figure 4.4 Trip Length Distribution of Trips Starting and Ending in Tampa FAF Zone

during (a) Weekdays (61,465 trips), and (b) Weekend (7,780 trips) .......................... 34

Figure 4.5 Trip Time Distribution of Trips Starting and Ending in Tampa FAF Zone


Figure 4.6 Trip Speed Distribution of Trips Starting and Ending in Tampa FAF Zone


Figure 4.7 Time-of-Day Profile of Trips Starting and Ending in Tampa FAF Zone


xvi

Figure 4.8 Destinations of Trips Starting in Hillsborough County (87,701 Trips) ....................... 38

Figure 4.9 Origins of Trips Ending in Hillsborough County (86,254 Trips) ................................ 39

Figure 4.10 Destinations of Trips Starting in Florida (1,102,873 Trips) ...................................... 40

Figure 4.11 Origins of Trips Ending in Florida (1,097,614 Trips) ............................................... 41

Figure 4.12 Distribution of Ping Rates from One Week of ATRI Data with and without

Spot Speeds (May 2010) ........................................................................................... 43

Figure 4.13 Route Choice for 365 Trips between a Sample OD Pair (3124-3703) ...................... 45

Figure 4.14 Comparison of OD Travel Times Measured in ATRI Data (excluding all

stops en route) with Travel Times from Google Maps ............................................. 48

Figure 4.15 Comparison of OD Travel Times Measured in ATRI Data (excluding all

stops en route) with Travel Times used in FLSWM ................................................. 49

Figure 5.1 Distribution of Trucks by Daily Trip Rate .................................................................. 51

Figure 5.2 Observed Heavy Truck Traffic Flows at Different Telemetered Traffic Counting

Sites (TTMS) in Florida .............................................................................................. 53

Figure 5.3 Heavy Truck (Classes 8–13) Counts from TTMS Data vs. Truck Counts from

ATRI Data during May 9-15, 2010............................................................................. 54

Figure 5.4 FLSWM TAZ-Level Trip Productions and Attractions in the ATRI Data

(4 Months of Data Factored to One Day) ................................................................... 56

Figure 5.5 County-Level Trip Productions and Attractions in the ATRI Data

(4 Months of Data Factored to One Day) ................................................................... 57

Figure 5.6 Percentage of Observed Heavy Truck (Classes 8–13) Volumes Represented

by ATRI Data at Telemetric Traffic Monitoring Sites in Florida, during

May 9–15, 2010 .......................................................................................................... 58

Figure 6.1 Schematic of the ODME Procedure Used in This Project .......................................... 61

Figure 6.2 Spatial Distribution of Telemetered Traffic Monitoring Sites (TTMS) in Florida

Used for ODME .......................................................................................................... 69

Figure 6.3 Spatial Distribution of Traffic Counting Stations in U.S. Used for ODME ............... 70

Figure 6.4 Observed vs. Estimated Heavy Truck Counts per Day at Different Locations

in Florida ..................................................................................................................... 76

xvii

Figure 6.5 Trip Length Distributions of Trips in Estimated and Seed OD Matrices ................... 77

Figure 6.6 Comparison of Trip Productions by County between Seed OD Matrix and

Estimated OD Matrix .................................................................................................. 78

Figure 6.7 Comparison of Trip Attractions by County between Seed OD Matrix and

Estimated OD Matrix .................................................................................................. 79

Figure 6.8 Comparison of Truck Trip Productions and Attractions between Seed and

Estimated OD Matrices ............................................................................................... 80

Figure 6.9 Comparison of Truck Trip Flows between Each OD Pair in the Seed and

Estimated OD Matrices ............................................................................................... 83

Figure A.1 Visualization of Truck Re-routing after Highway (I-75) Closure due to a

Multi-vehicle Crash ................................................................................................... 94

Figure A.2 Analysis of Truck Flows from Blount Island ............................................................. 95

Figure A.3 Truck Trips within 1 Hour of Departing Blount Island .............................................. 96

Figure A.4 Urban Truck Trips within 1 Hour of Departing Blount Island ................................... 96

Figure A.5 Analysis of Truck Flows from Port Everglades ......................................................... 97

Figure A.6 Truck Trips within 1 Hour of Departing Port Everglades .......................................... 98

Figure A.7 Urban Truck Trips within 1 Hour of Departing Port Everglades ............................... 98

Figure A.8 Travel Routes of Trucks Crossing the Traffic Counting Station on I-75 at Ocala ..... 99

Figure A.9 Routing Patterns of Southbound Trucks Using US 301 in Florida .......................... 100

Figure A.10 Routing Patterns of Northbound Trucks Using US 301 in Florida ........................ 100

Figure B.1 Trip Length Distribution of Trips Starting and Ending in Tampa FAF Zone

during (a) Weekdays (61,465 trips), and (b) Weekend (7,780 trips) ...................... 102

Figure B.2 Trip Time Distribution of Trips Starting and Ending in Tampa FAF Zone


Figure B.3 Trip Speed Distribution of Trips Starting and Ending in Tampa FAF Zone


Figure B.4 Time-of-Day Profile of Trips Starting and Ending in Tampa FAF Zone


xviii

Figure B.5 Destinations of Trips Starting in Hillsborough County (87,701 Trips) .................... 106

Figure B.6 Origins of Trips Ending in Hillsborough County (86,254 Trips) ............................. 107

Figure B.7 Trip Length Distribution of Trips Starting and Ending in Orlando FAF Zone

during (a) Weekdays (79,094 trips), and (b) Weekend (10,528 trips) .................... 108

Figure B.8 Trip Time Distribution of Trips Starting and Ending in Orlando FAF Zone


Figure B.9 Trip Speed Distribution of Trips Starting and Ending in Orlando FAF Zone


Figure B.10 Time-of-Day Profile of Trips Starting and Ending in Orlando FAF Zone

during (a) Weekdays (79,094 trips), and (b) Weekend (10,528 trips) ................... 111

Figure B.11 Destinations of Trips Starting in Orange County (94,575 Trips) ........................... 112

Figure B.12 Origins of Trips Ending in Orange County (92,994 Trips) .................................... 113

Figure B.13 Trip Length Distribution of Trips Starting and Ending in Jacksonville FAF Zone

during (a) Weekdays (56,319 trips), and (b) Weekend (5,789 trips) ..................... 114

Figure B.14 Trip Time Distribution of Trips Starting and Ending in Jacksonville FAF Zone


Figure B.15 Trip Speed Distribution of Trips Starting and Ending in Jacksonville FAF Zone


Figure B.16 Time-of-Day Profile of Trips Starting and Ending in Jacksonville FAF Zone


Figure B.17 Destinations of Trips Starting in Duval County (110,314 Trips) ........................... 118

Figure B.18 Origins of Trips Ending in Duval County (111,023 Trips) .................................... 119

Figure B.19 Trip Length Distribution of Trips Starting and Ending in Miami FAF Zone


Figure B.20 Trip Time Distribution of Trips Starting and Ending in Miami FAF Zone


Figure B.21 Trip Speed Distribution of Trips Starting and Ending in Miami FAF Zone


xix

Figure B.22 Time-of-Day Profile of Trips Starting and Ending in Miami FAF Zone


Figure B.23 Destinations of Trips Starting in Miami-Dade County (70,273 Trips)................... 124

Figure B.24 Origins of Trips Ending in Miami-Dade County (69,274 Trips) ............................ 125

Figure B.25 Trip Length Distribution of Trips Starting and Ending in Other Regions in

Florida during Weekdays (527,484 Trips), and (b) Weekend (73,678 Trips) ....... 126

Figure B.26 Trip Time Distribution of Trips Starting and Ending in Other Regions in


Figure B.27 Trip Speed Distribution of Trips Starting and Ending in Other Regions in


Figure B.28 Time-of-Day Profile of Trips Starting and Ending in Other Regions in Florida

during (a) Weekdays (527,484 Trips), and (b) Weekend (73,678 Trips) .............. 129

Figure B.29 Destinations of Trips Starting in Polk County (133,365 Trips) .............................. 130

Figure B.30 Origins of Trips Ending in Polk County (132,393 Trips) ....................................... 131

Figure B.31 Trip Length Distribution of Trips Starting and Ending in Florida during

(a) Weekdays (808,673 trips), and (b) Weekend (106,033 trips) .......................... 132

Figure B.32 Trip Time Distribution of Trips Starting and Ending in Florida during


Figure B.33 Trip Speed Distribution of Trips Starting and Ending in Florida during


Figure B.34 Time-of-Day Profile of Trips Starting and Ending in Florida during


Figure B.35 Trip Length Distribution of Trips Starting in Florida and Ending outside

Florida during (a) Weekdays (138,816 trips), and (b) Weekend (13,900 trips) .... 136

Figure B.36 Trip Time Distribution of Trips Starting in Florida and Ending outside


Figure B.37 Trip Speed Distribution of Trips Starting in Florida and Ending outside


Figure B.38 Time-of-Day Profile of Trips Starting in Florida and Ending outside Florida

during (a) Weekdays (138,816 trips), and (b) Weekend (13,900 trips) ................. 139

xx

Figure B.39 Destinations of Trips Starting in Florida (1,102,873 Trips) ................................... 140

Figure B.40 Trip Length Distribution of Trips Starting outside Florida and Ending in


Figure B.41 Trip Time Distribution of Trips Starting outside Florida and Ending in


Figure B.42 Trip Speed Distribution of Trips Starting outside Florida and Ending in


Figure B.43 Time-of-Day Profile of Trips Starting outside Florida and Ending in Florida

during (a) Weekdays (127,796 trips), and (b) Weekend (15,635 trips) ................. 144

Figure B.44 Origins of Trips Ending in Florida (1,097,614 Trips) ............................................ 145

Figure C.1 Route Choice for 365 Trips for OD Pair “3124-3703” ............................................. 147


Figure C.3 Route Choice for 327 Trips between OD Pair “2228-4227” .................................... 148


Figure C.5 Route Choice for 386 Trips for OD Pair “5983-792” ............................................... 149


Figure C.7 Route Choice for 28 Trips for OD Pair “4035-5819” ............................................... 150

Figure C.8 Route Choice 38 Trips between OD Pair “5073-6117” ............................................ 150

Figure C.9 Route Choice for 23 Trips for OD Pair “413-6086” ................................................. 151

Figure C.10 Route Choice for 28 Trips between OD Pair “2355-6176” .................................... 151

Figure C.11 Travel Time Distribution for Trips between OD Pair “3124-3703” (384 Trips) .... 152

Figure C.12 Trips between OD Pair “3124-3703” during Weekdays (284 Trips) ..................... 152

Figure C.13 Mid-day Trips between OD Pair “3124-3703” during Weekdays (111 Trips) ....... 153

Figure C.14 PM Peak Trips between OD Pair “3124-3703” during Weekdays (88 Trips) ........ 153

Figure C.15 Night-time Trips between OD Pair “3124-3703” during Weekdays (83 Trips) ..... 154

xxi




Figure C.19 Travel Time Distribution for Trips between OD Pair “2228-4227” during

Weekdays (341 Trips) ............................................................................................ 157

Figure C.20 Night-time Trips between OD Pair “2228-4227” during Weekdays (166 Trips) ... 157

Figure C.21 AM Peak Trips between OD Pair “2228-4227” during Weekdays (173 Trips) ..... 158



Figure C.24 AM Peak Trips between OD Pair “3662-3124” during Weekdays (30 Trips) ....... 160

Figure C.25 Mid-day Trips between OD Pair “3662-3124” during Weekdays (54 Trips) ......... 160


Figure C.27 Travel Time Distribution for Trips between OD Pair “5983-792” (405 Trips) ...... 162

Figure C.28 Trips between OD Pair “5983-792” during Weekdays (398 Trips) ....................... 162

Figure C.29 AM Peak Trips between OD Pair “5983-792” during Weekdays (132 Trips) ....... 163



Figure C.32 Trips between OD Pair “2420–4147” during Weekdays (166 Trips) ..................... 164

Figure C.33 Night-time Trips between OD Pair “2420–4147” during Weekdays (162 Trips) .. 165


Weekdays (29 Trips) .............................................................................................. 166


Figure C.36 Travel Time Distribution for Trips between OD Pair “5073-6117” (39 Trips) ...... 167

Figure C.37 Trips between OD Pair “5073-6117” during Weekdays (33 Trips) ....................... 167

xxii



Weekdays (24 Trips) .............................................................................................. 169

Figure C.40 Night-time Trips between OD Pair “413-6086” during Weekdays (15 Trips) ....... 169


Weekday (28 Trips) ............................................................................................... 170

Figure C.42 Mid-day Trips between OD Pair “2355-6176” during Weekdays (19 Trips) ......... 170

Figure D.1 Florida Truck Flows (Counts for Single Unit Trucks) for Facility Type:

Urban Freeway ........................................................................................................ 173


Other Freeway ......................................................................................................... 174


Controlled Access Expressway ............................................................................... 175


Divided Arterial Unsignalized ................................................................................. 176


Divided Arterial Class 1a ........................................................................................ 177


Undivided Arterial Unsignalized with Turn Bays ................................................... 178


Major Local Divided Roadway ............................................................................... 179


Other Local Undivided Roadway without Turn Bays ............................................. 180


Freeway Group 1 HOV Lane (Non-Separated) ....................................................... 181


Freeway Group 1 Toll Facility ............................................................................... 182


Other Freeway Toll Facility ................................................................................... 183

Figure D.12 Florida Truck Flows (Counts for Single Unit Trucks) for Area Type:

Central Business District........................................................................................ 185

xxiii


Residential Area of Urbanized Areas .................................................................... 186


Undeveloped Portions of Urbanized Areas............................................................ 187


Transitioning Areas/Urban Areas over 5,000 Population ...................................... 188


Other Outlying Business District ........................................................................... 189


Developed Rural Areas/Small Cities under 5,000 Population ............................... 190


Undeveloped Rural Areas ...................................................................................... 191

Figure D.19 Florida Truck Flows (Counts for Tractor-Trailer Trucks) for Facility Type:

Urban Freeway Group 1......................................................................................... 193


Other Freeway ........................................................................................................ 194


Controlled Access Expressway .............................................................................. 195


Divided Arterial Unsignalized ............................................................................... 196


Divided Arterial Class 1a ....................................................................................... 197


Undivided Arterial Unsignalized with Turn Bays ................................................. 198


Major Local Divided Roadway .............................................................................. 199


Other Local Undivided Roadway without Turn Bays ........................................... 200


Freeway Group 1 HOV Lane ................................................................................. 201

xxiv


Freeway Group 1 Toll Facility ............................................................................... 202


Other Freeway Toll Facility ................................................................................... 203

Figure D.30 Florida Truck Flows (Counts for Tractor-Trailer Trucks) for Area Type:

Central Business District........................................................................................ 205


Residential Areas of Urbanized Areas ................................................................... 206


Undeveloped Portions of Urbanized Areas............................................................ 207


Transitioning Areas/Urban Areas over 5,000 Population ...................................... 208


Other Outlying Business District ........................................................................... 209


Developed Rural Areas/Small Cities .................................................................... 210


Undeveloped Rural Areas ...................................................................................... 211

Figure D.37 a) Distribution of Daily Truck Counts in Traffic Sites; b) Same Distribution by

Facility Type .......................................................................................................... 212

xxv

LIST OF TABLES

Table 2.1 Truck Speed Reliability Measures for Eight Miami Corridors..................................... 14

Table 3.1 Distribution of Time Gap between GPS Readings in a Week of

Truck GPS Data ............................................................................................................ 18

Table 3.2 Summary of Truck Trips Extracted from Four Months of ATRI’s

Truck GPS Data ............................................................................................................ 29

Table 4.1 Distribution of OD Pairs Based on Number of Trips Extracted from

ATRI Data .................................................................................................................... 44

Table 4.2 Travel Time and Travel Distance Measurements for Sample of 10 OD Pairs ............. 47

Table 5.1 Aggregate Coverage of Heavy Truck Traffic Volumes in Florida by

ATRI Data (for One week from May 9–15, 2010) ....................................................... 55

Table 6.1 County-to-County OD Pairs in Florida with No Trips in the Seed OD Flow Matrix

Derived from ATRI’s Truck GPS Data ........................................................................ 64

Table 6.2 Florida County to Non-Florida State OD Pairs with No Trips in the Seed OD Flow

Matrix Derived from ATRI’s Truck GPS Data ............................................................ 66

Table 6.3 RMSE Values between Estimated and Observed Heavy Truck Traffic Volumes at

Different Locations in Florida for Different Assumptions in ODME Procedure ......... 72

Table 6.4 Summary of Truck Trips in Seed and Estimated OD Matrices .................................... 74

Table 6.5 Heavy Truck Trip Flows from Florida to Other States (Greater than 0.5 % in

Estimated Matrix) ......................................................................................................... 82

Table 6.6 Heavy Truck Trip Flows from Other States to Florida (Greater than 0.5 % in

Estimated Matrix) ......................................................................................................... 82

Table 6.7 Destinations of Heavy Truck Trip Flows from Selected Counties in Florida .............. 82

Table 6.8 Origins of Heavy Truck Trip Flows to Selected Counties in Florida ........................... 83

Table D.1 Dates Forming Each Season ...................................................................................... 171

Table D.2 Seasonality Analysis of Single Unit Trucks by Facility Type ................................... 172

Table D.3 Description of Facility Types (FTYPE) ..................................................................... 172

Table D.4 Seasonality Analysis of Single-Unit Trucks by Area Type ....................................... 184

xxvi

Table D.5 Description of Area Types (ATYPE) ........................................................................ 184

Table D.6 Seasonality Analysis of Tractor-Trailer Trucks by Facility Type ............................. 192

Table D.7 Seasonality Analysis of Tractor-Trailer Trucks by Area Type .................................. 204

1

CHAPTER 1 : INTRODUCTION

1.1 Background

Freight is gaining increasing importance in transportation planning and decision making at all

levels of the government, particularly MPOs, states, and at the federal level. An accelerated

growth in the volume of freight shipped on American highways has led to a significant increase

in truck traffic. This has put enormous pressure on national highways, impacting traffic

operations, safety, highway infrastructure, port operations, and distribution center operations.

Traffic congestion, in turn, impedes the speed and reliability of freight movement on the

highway system and leads to direct economic costs for producers and consumers, passenger

traffic congestion, safety issues, and environmental impacts.

As freight movement continues to grow within and between urban areas, appropriate

planning and decision making processes are necessary to mitigate the above-mentioned impacts.

However, a main challenge in establishing these processes is the lack of adequate data on freight

movements such as detailed origin-destination (OD) data, truck travel times, freight tonnage

distribution by OD pairs, transported commodity by OD pairs, and details about truck trip stops

and paths. As traditional data sources on freight movement are either inadequate or no longer

available (e.g., the Vehicle Inventory and Use Survey), new sources of data must be investigated.

Recognizing the need for better freight data, several efforts are underway to exploit

advanced technologies and form innovative partnerships with freight stakeholders to gather

freight movement data. Many trucking companies use advanced vehicle monitoring (AVM)

systems that allow remote monitoring of their fleets using Geographical Positioning Systems

(GPS) technology-based Automatic Vehicle Location (AVL) systems. To tap into such truck

GPS data sources, private-sector truck data providers have formed innovative partnerships with

freight carriers and other freight stakeholders to collect the GPS data and provide it to public

agencies while protecting the confidentiality of the data. Notable among such efforts is a joint

venture by the American Transportation Research Institute (ATRI) and the Federal Highway


performance measures (FPM) on key corridors in the nation. This FPM data system is built based

on data obtained from trucking companies that use GPS-based AVM/AVL technologies to

remotely monitor their trucks. ATRI’s FPM database contains GPS traces of a large number of

trucks as they traveled through the national highway system. This provides unprecedented

amounts of data on freight truck movements throughout the nation (and Florida). Such truck GPS

data potentially can be used to support planning, operation, and management processes

associated with freight movements. Further, the data can be put to better use when used in

conjunction with other freight data obtained from other sources.

ATRI’s truck GPS data is being used increasingly for a variety of freight performance

measurement and planning applications in the recent past. The applications include, but are not

limited to, identifying and prioritizing major freight bottlenecks on the nation’s highways (Short

et al., 2009) and regional and statewide truck flow modeling (Bernardin et al., 2011; Kuppam et

al., 2014). For the state of Florida, ATRI’s truck GPS provides a significant opportunity to

develop data on statewide truck flow patterns, freight performance measurement, and a variety of

other applications. Since a majority of freight being shipped across the nation (and Florida) is via

2

the truck mode, ATRI’s truck GPS data are likely to be of significant value in providing the data

needed for understanding statewide freight movement by the truck mode.

The Florida Department of Transportation (FDOT) is developing a program to develop

data on and understand statewide freight movements, including the freight flows between

different origins and destinations in the state, the freight flows coming into and going out of the

state, the performance of the transportation system in accommodation freight movement, and

freight flows into and out of major freight activity centers in the state. As part of this program,

FDOT uses FHWA’s freight analysis framework (FAF) and other data sources to understand

current and future commodity flows in the state. For example, FDOT is developing methods to

disaggregate FAF data to obtain commodity flows at smaller geographies such as counties and

census tracts. The FDOT program hopes to relate such disaggregate commodity flow data to the

information gathered from the ATRI data on truck movements to understand how these

commodities are transported between different origins and destinations. In another ongoing

project, FDOT is developing a next-generation multimodal freight forecasting model system for

long-term freight planning in Florida. FDOT intends to validate and calibrate this model with an

independent source of data on OD truck flows in the state.

1.2 Project Objectives

The overarching goal of this project was to investigate the use of ATRI’s truck GPS data for

statewide freight performance measurement, statewide freight truck flow analysis, and other

freight planning and modeling applications. The project aimed to develop supporting methods

and procedures to use the data, and combine the data with other data sources, for different freight

movement modeling and planning applications. The specific objectives of the project are

identified next.

1.2.1 Objective 1: Derive Freight Performance Measures for Florida’s

Freight-intensive Highways

The first objective of the project was to use ATRI’s truck GPS data to derive freight performance

measures for Florida’s freight-intensive highways. To this end, the project involved the

development of data on average truck speeds on each (and every) mile of Florida’s Strategic

Intermodal System (SIS) highway network for different time periods in the day. These data were

developed and submitted in a GIS shape file that can be used in an ArcGIS environment to

identify the major freight bottlenecks on Florida’s SIS highway network. In addition to the

development of average speed measures, the project developed example applications of ATRI’s

truck GPS data for measuring truck speed reliability and for highway freight bottleneck analysis.

1.2.2 Objective 2: Develop Methods to Convert ATRI’s Raw GPS Data Streams

into Truck Trips


full potential of the data for freight planning applications. Therefore, the second objective of the

project was to convert the raw GPS data into a data base of truck trips. Development of such a

truck trip database involved the determination of truck starting and ending instances and

locations, trip distance, total trip duration, and duration of intermediate stops (e.g., at traffic

signals and rest stops) in the trip. In addition, the process involved resolution of potential

3

anomalies in GPS data, such as data-discontinuities due to loss of satellite signals. This task

involved the development of algorithms and a software code (written in Java programming

language) to convert the raw GPS data streams into a truck trip format. These algorithms were

then applied to four months of raw GPS data from ATRI, comprising a total of 145 million raw

GPS data records, to develop a large database of truck trips traveling within, into, and out of the

state. The resulting database comprised more than 1.2 million truck trips traveling within, into,

and out of the state.

1.2.3 Objective 3: Analyze Truck Travel Characteristics in Florida

The third objective of the project was to use ATRI’s truck GPS data to analyze truck travel

characteristics in the state of Florida. To this end, the project involved an analysis of the truck

trip data derived from the four months of ATRI’s truck GPS data. The truck travel characteristics

analyzed included trip duration, trip length, trip speed, time-of-day profiles, and OD flows. Each

of these characteristics was derived at a statewide level and for different regions in the state—

Jacksonville, Tampa Bay, Orlando, Miami, and rest of Florida—defined based on the freight

analysis framework (FAF) zoning system. Furthermore, the task involved deriving and analyzing

the OD travel distances, travel times, and travel routes between a selected set of OD pairs in the

state.

1.2.4 Objective 4: Assess ATRI’s Truck GPS Data and Its Coverage of Truck Traffic

in Florida

The fourth objective was to assess ATRI’s truck GPS data in Florida to gain an understanding of

its coverage of truck traffic in the state of Florida. This included deriving insights on (a) the

types of trucks (e.g., heavy trucks and light trucks) present in the data, and (b) the geographical

coverage of the data in Florida, and (c) the proportion of the truck traffic flows in the state

covered by the data.

1.2.5 Objective 5: Derive Statewide Truck Trip Flow OD Tables for the

Traffic Analysis Zone (TAZ)-level Spatial Resolution in the

Florida Statewide Model (FLSWM)

An important objective of the project was to use ATRI’s truck GPS data in combination with

other available data sources to derive origin-destination (OD) tables of freight truck flows within,

into, and out of the state of Florida. As part of this task, first, the truck trip database developed

from four months of ATRI’s GPS data was converted into OD tables at the TAZ-level spatial

resolution used in the Florida Statewide Model (FLSWM). Such an OD table derived only from

the ATRI data, however, was not necessarily representative of the freight truck flows in the state.

This is because the ATRI data did not comprise the census of trucks in the state; the data

comprised only a sample of trucks traveling in the state. Although it is a large sample, it is not

necessarily a random sample and is likely to have spatial biases in its representation of truck

flows in the state. To address these issues, the OD tables derived from the ATRI data need to be

combined with observed truck traffic volumes at different locations in the state (and outside the

state) to derive a more robust OD table that is representative of the freight truck flows within,

into, and out of the state. To achieve this, a mathematical procedure called the Origin-

Destination Matrix Estimation (ODME) method was employed to combine the seed OD flow

4

matrix generated from the ATRI data with observed truck traffic volume information at different

locations within and outside Florida.

1.2.6 Objective 6: Explore the Use of ATRI’s Truck GPS Data for Other Applications

of Interest to Florida

Another objective of the project was to explore the use of ATRI’s truck GPS data for a variety of

applications of interest to Florida other than those mentioned above. Specifically, exploratory

analysis tasks were provided on an ongoing basis as needed by FDOT. These included (a)

analysis of truck flows out of two ports in Florida—Port Blount Island in Jacksonville and Port

Everglades in Fort Lauderdale, (b) analysis of routing patterns of trucks that used the US 301

roadway to travel between I-95 around Jacksonville and I-75 around Ocala, and (c) analysis of

changes in truck routing patterns during the closure of a stretch of I-75 near Ocala due to a major

multi-vehicle crash in January 2012. Note here that these applications are only exploratory in

nature and conducted only as proofs of concept. Future work can expand on these explorations to

conduct full-scale applications of the above mentioned tasks.

1.3 Structure of the Report

The research conducted for the above-mentioned objectives and the project outcomes and

findings are described in subsequent chapters.

Chapter 2 presents a brief overview of ATRI’s truck GPS data. In addition, the chapter

describes the procedures and outcomes of the freight performance measurement task described in

objective 1. The chapter is accompanied by an appendix (Appendix A) to provide an overview of

the exploratory tasks described in objective 6.

Chapter 3 presents the procedures and algorithms used to convert the raw GPS data

streams into a large database of truck trips in Florida (objective 3).

Chapter 4 presents an analysis of the truck travel characteristics using the truck trip

database developed from four months of raw GPS data (objective 4). This chapter is

accompanied by an appendix to present examples of truck route choice patterns between several

OD pairs as well as distributions of truck travel times between those same OD pairs.

Chapter 5 presents an assessment of ATRI’s truck GPS data in Florida, specifically in

terms of its coverage of truck traffic flows in the state of Florida as well as different geographical

locations (objective 5). This chapter is accompanied by an appendix that presents an analysis of

the seasonality of the observed truck traffic counts obtained from the Telemetered Traffic

Monitoring Sites (TTMS) data collected by FDOT.

Chapter 6 presents the development of statewide OD truck flows using ATRI data in

combination with other data sources for the TAZ-level spatial resolution in the FLSWM. The

ODME procedure used for this purpose, the observed truck traffic volume data within and

outside Florida, and the results from the procedure are presented in this chapter.

Chapter 7 summarizes the work conducted, the findings and benefits of the study, and

identifies opportunities for future research and implementation.

5

CHAPTER 2 : OVERVIEW OF ATRI DATA AND ITS APPLICATIONS FOR

FREIGHT PERFORMANCE MEASUREMENT

2.1 Introduction

This chapter provides a background on ATRI’s truck GPS data. In addition, the chapter provides

examples of the applications of the data for highway freight performance measurement (section

2.2), with a focus on the performance measures (i.e., average truck speeds) developed for

Florida’s Strategic Intermodal System (SIS) highway network.

2.2 Background on ATRI's Truck GPS Data

The American Transportation Research Institute (ATRI) collects truck position data throughout

the U.S. and North America from a large sample of trucks that use onboard, wireless

communications systems. This information has been collected for FHWA as part of their Freight

Performance Monitoring System (FPMS) (Jones et al., 2005) for the purpose of monitoring truck

travel speeds in freight-significant corridors. The ATRI database contains consecutive truck GPS

data for several locations in the nation from 2002 through the most recent month of 2014. In the

state of Florida, the ATRI database contains truck GPS data from 2005 through the most recent

month of 2014. For the majority of analysis conducted in this project, ATRI used data in Florida

from the calendar year 2010.

At a minimum each record within the database contained the following information:

Unit Information: A unique identifier for the transponder/truck,

Geographic Information: The latitude and longitude data that identify where a truck

position record was recorded, and

Temporal Information: The time at which a truck position record was recorded, in the

following format – MM-DD-YYYY HH:MM:SS.

Additionally, many of the records contain information such as spot speed and heading.

Spot speed is the “instantaneous” speed of the truck at the location where its movement is

recorded. A sample record looks as below:

Unique Truck ID Time/Date Stamp Latitude Longitude Heading Speed

12232123 05-03-2011 01:55:55 33.915932 -84.494760 N 25

Note that the “Unique Truck ID” is a random digit identifier that cannot be used to

identify the actual vehicle or to trace the carriers that provided the data. The original truck GPS

identification numbers were converted to random digit identifiers and destroyed to protect the

confidentiality of the carriers. Therefore, ATRI is unable to provide information about

individuals trucks (other than GPS records), such as the commodity carried, weight or volume

carried, purpose of travel, and the type of truck. However, since the focus of FPMS was truck

travel speeds on freight-significant corridors, and the trucks traveling in these corridors tend to

be predominantly tractor-trailer combination trucks, ATRI indicated that the data comprise

predominantly tractor-trailer combination trucks (also called heavy trucks from here forward).

Most such trucks can be categorized as classes 8–13 of FHWA’s vehicle classification scheme,

6

as shown in Figure 2.1. It will be discussed in Chapter 5 that an unknown but small portion of

the data comprise trucks of lower classification (Class 7 or below), such as single unit trucks that

might serve the purpose of freight distribution in urban areas and across smaller distances.

To protect confidentiality of the data, ATRI required USF to sign a non-disclosure

agreement (NDA). According to the NDA, the raw GPS data shared by ATRI with USF was to

be used only for the purpose of analysis by USF researchers. The raw data must not be shared

with anyone outside the research team. However, the agreement allows for the aggregate results

and data products from the research to be submitted to FHWA as long as the locations of the trip

origins, destinations, and intermediate locations are not revealed in a high spatial resolution.

Once the NDA was in place, ATRI started sharing the raw truck GPS data with USF. The data

were shared through a secure FTP site that was used throughout the project for transferring data.

In this project, the research team worked with Florida-specific data for the year 2010.

Specifically, most of the tasks were carried out using ATRI’s truck GPS data for a certain time

period in the year 2010. For example, for generating of average truck travel speeds on the Florida

network, three months of data in the year 2010 were used. For other tasks, such as generating OD

flows of truck travel in the state, four months of data were used.

For any given time period (say, a week), the first step in the process of extracting a

Florida dataset involved isolating all the unique truck IDs with GPS positions in Florida during

that time period (i.e., all trucks found in Florida during that period). Subsequently, all the GPS

data for those unique IDs were extracted, not only within the boundaries of Florida but also

throughout North America. Thus, the database can be used to derive information on truck travel

within Florida as well as that coming into and going out of Florida. For example, Figure 2.2

shows a sample of truck GPS records in Florida found during a certain time period in ATRI’s

database. Figure 2.3 shows all of ATRI’s truck GPS data in North America with the same unique

truck IDs found in Florida during that time period (i.e., the truck IDs in Florida from the data in

Figure 2.2).

Since analyzing large streams of data was complex and time consuming, most procedures

developed in the project were first developed for smaller-size datasets —specifically, data for

one week time periods—and then employed (and if needed modified) for larger datasets.

7

Figure 2.1 FHWA Vehicle Classifications

Source: http://onlinemanuals.txdot.gov/txdotmanuals/tri/images/FHWA_Classification_Chart_FINAL.png, accessed on 5-13-2014.

http://onlinemanuals.txdot.gov/txdotmanuals/tri/images/FHWA_Classification_Chart_FINAL.png

8

Figure 2.2 Sample of Truck GPS Record Positions in Florida

9

Figure 2.3 Truck GPS Record Positions throughout North America for

Same Sample of Trucks in Florida from Figure 2.2

2.3 Applications of ATRI’s Truck GPS Data for Freight Performance Measurement

2.3.1 Freight Performance Measures: Average Truck Speeds on SIS Highways

The ATRI dataset can be processed into measurements of average speeds for freight-significant

corridors. In this project, ATRI’s truck GPS data in Florida were used to measure average speeds

on each mile of Florida’s SIS highway network for different time periods of the day. Three

months of data in the year 2010 were used to generate the speeds.

The first step in generating the average truck speed information for SIS highways was to

identify truck GPS data points that fell on the SIS highway network segments. This required a

GIS layer of highly accurate and a bidirectional roadway network (i.e., a network that recognizes

divided highways—for example, a section of interstate 4 between Tampa and Orlando, as two

different segments, one in each direction) to select only those GPS data points that fell on the SIS

highway network. Extracting data that was not on the network (for example, parked at a facility

near the network) could produce less-reliable results.

10

The research team received from FDOT two different sources of network geometry in a

shape file format—(1) SIS network, and (2) Navteq network. After preliminary assessment by

the research team, it was determined that neither network was accurate enough for the purpose of

this task.1 Therefore, using these two networks as reference and by tracing the truck GPS points

on selected roadways, the ATRI team created a highly-accurate network geometry of Florida’s

SIS highways. The shape file of this network has a vector spatial representation and the

geographic coordinate reference GCS_WGS_1984. The network geometry data were organized

into roadway segments that were generally 1.0 miles in length in rural areas and 0.5 miles in

length in urban areas. The final network geometry comprised 9,750 individual segments covering

9,042 miles of multidirectional roadways on SIS.

Next, the truck GPS dataset was further narrowed to include only those data points that

fell along Florida’s SIS road network (using the ATRI-created network geometry). Using the

geographic coordinates of the truck GPS points, each truck data point was assigned a unique

segment ID that linked each point to a specific highway facility. A spatial join of the truck GPS

dataset to the network accomplished this task. The truck GPS dataset was processed into

measurements of average speed. With the network divided into segments and the truck GPS data

assigned a unique segment ID, the data for each segment were aggregated and sorted into the

following five time bins:

AM Peak: 6:00 a.m. – 9:59 a.m.

Mid-day: 10:00 a.m. – 2:59 p.m.

PM Peak: 3:00 p.m. – 6:59 p.m.

Off-peak: 7:00 p.m. – 5:59 a.m.

Average of all hours: 12 a.m. – 11:59 p.m.

Average speeds were then calculated for each corridor segment during each time period.

ATRI used several additional sources of data, such as maximum speed limits and Average

Annual Daily Truck Traffic (AADTT) to validate the average speed FPMs. The resulting shape

file, with information on average weekday speeds on each mile of the SIS highway network for

different time periods of the day, was submitted to FDOT. The shape file can be used in a GIS

workspace for visualizing average truck speeds and identifying locations affected by congestion.

As an illustration, Figure 2.4 shows the average AM Peak speed for each segment in the SIS

highway network based on data from weekdays in 2010. Segments with slower average speeds

are shown in shades of red, and those with faster average speeds are shown in shades of green.

Figure 2.5 displays the shape file and its functionality when deployed in ESRI’s ArcGIS desktop.

In this example, one segment is highlighted, and an information box displays the unique

attributes for that individual segment.

1 The SIS network provided by FDOT was not multidirectional, so it could not be used for measuring speeds by

direction. The Navteq network was multidirectional. However, at some locations, the network geometry was not

aligned with the true location of the roadway.

11

Figure 2.4 Average Truck Speeds for AM Peak Time Period on Florida’s SIS Highways

12

Figure 2.5 Average Truck Speeds for Different Time Periods in the Day for a

One-Mile Segment in the Miami Region

The performance measurements (i.e., average truck speeds), if produced for each hour,

can be used to analyze specific roadway segments and networks during a given time period. As

an illustration, Figure 2.6 displays average speeds in one-hour time periods across one-mile

segments along a stretch of I-95 in Fort Lauderdale. Such information can be used to understand

the spatial and temporal extent of congestion faced by trucks on freight significant corridors.

13

Figure 2.6 Average Speeds by Hour for Miles 24-48 on I-95 in Fort Lauderdale, Florida

2.3.2 Freight Performance Measures: Truck Speed Reliability Measurements

Since the ATRI data contain truck GPS data for extended time periods, the data can be used to

measure the reliability of truck movement (i.e., to develop measures of variability in truck

speeds) on freight-significant corridors in the state. Two widely-used measures of reliability

include the Travel Time Index (TTI) and Planning Time Index (PTI). The TTI is the ratio of the

average speed for a given highway segment at a particular time of day to the functional free-flow

speed of the same segment of highway. The higher the index, the longer it takes, on average, to

travel the segment of highway at that time of day compared to if traffic was moving at free flow.

The PTI is the ratio of the “worst-case scenario” average travel speed for a given highway

segment at a particular time of day to the functional free-flow speed of the same segment of

highway. The worst-case scenario is defined as the 95th percentile of average speeds during a

given time period (with lower speeds being a higher percentile rank). The higher the index, the

more time is needed for motorists to plan for a 95 percent on-time arrival.

Eight urban freeway corridors from southeast Florida (Figure 2.7) were selected for

reliability (of truck speeds) measurement demonstration in this study. Table 2.1 presents

reliability measures along each of these eight urban segments, presented separately by direction.

It can be observed from the table that the reliability was lowest during the PM peak periods.

14

Figure 2.7 Highway Corridors Selected for Truck Speed Reliability Measurements

Table 2.1 Truck Speed Reliability Measures for Eight Miami Corridors

2.3.3 Identification and Analysis of Freight Bottlenecks

Identification and prioritization of highway bottlenecks offer a better understanding of when,

where, and possibly why congestion is occurring. As an illustration, Figure 2.8 provides an

example where the freight performance measures (average truck speeds) are fused with the FAF

data on truck volumes to identify important bottlenecks for freight movement along Florida’s

interstate highways. Specifically, the color coding displays average speeds on interstate

M i a m i , F L - A p r i l - J u n e 2 0 1 0

P o ly lin e

F re e w a y S e c tio n

(s o rte d fro m m o s t c o n g e s te d to

le a s t c o n g e s te d )

L e n g th

(m i)

M o rn in g P e a k

(6 a -9 a )

M id d a y

(9 a -4 p )

E v e n in g

P e a k

(4 p -7 p )

A v e ra g e

P e a k

P e rio d

M o rn in g

P e a k

(6 a -9 a )

M id d a y

(9 a -4 p )

E v e n in g

P e a k

(4 p -7 p )

A v e ra g e

P e a k

P e rio d

I-9 5 N B : S R 9 1 to S R 8 4 1 2 .2 1 .1 6 1 .1 0 1 .1 8 1 .1 7 1 .6 9 1 .3 5 2 .0 4 1 .8 2

I-9 5 S B : S R 8 4 to S R 9 1 1 2 .2 1 .1 2 1 .1 6 1 .3 6 1 .1 9 1 .6 8 1 .7 7 3 .1 2 2 .1 2

I-9 5 N B : I-1 9 5 to S R 9 1 8 .2 1 .0 9 1 .1 7 2 .0 7 1 .6 8 1 .3 8 2 .4 0 1 0 .0 0 6 .5 6

I-9 5 S B : S R 9 1 to I-1 9 5 8 .2 1 .3 6 1 .2 7 1 .3 3 1 .3 6 3 .4 1 2 .9 3 4 .1 6 3 .4 9

I-9 5 N B : U S 1 to I-1 9 5 5 .1 1 .2 1 1 .1 8 1 .6 5 1 .3 8 3 .1 7 1 .6 4 4 .3 1 3 .6 2

I-9 5 S B : I-1 9 5 to U S 1 5 .1 1 .2 5 1 .1 8 1 .3 9 1 .2 7 3 .1 3 2 .2 4 4 .5 7 3 .3 1

I-7 5 N B : S R 8 2 1 to I-5 9 5 1 3 .3 1 .0 4 1 .0 5 1 .1 5 1 .0 6 1 .1 1 1 .1 5 1 .6 6 1 .2 1

I-7 5 S B : I-5 9 5 to S R 8 2 1 1 3 .3 1 .0 9 1 .0 6 1 .0 9 1 .0 9 1 .3 5 1 .1 6 1 .2 8 1 .3 3

S R 8 2 6 N B : S R 8 3 6 to I-7 5 8 .3 1 .1 2 1 .1 8 1 .8 6 1 .5 3 1 .5 0 1 .6 7 5 .2 0 3 .5 6

S R 8 2 6 S B : I-7 5 to S R 8 3 6 8 .3 1 .2 8 1 .3 1 1 .5 1 1 .3 2 2 .1 8 2 .5 7 7 .5 9 3 .1 0

S R 9 1 N B : S R 8 2 1 to S R 8 5 2 2 .9 1 .3 5 1 .1 8 1 .2 9 1 .3 3 3 .1 5 2 .3 0 3 .9 7 3 .3 6

S R 9 1 S B : S R 8 5 2 to S r 8 2 1 0 .3 1 .2 9 1 .2 1 1 .1 7 1 .2 6 5 .9 6 4 .1 5 1 .8 1 4 .9 3

I-7 5 N B : S R 8 2 6 to S R 8 2 1 4 .9 1 .0 5 1 .0 5 1 .0 8 1 .0 6 1 .1 0 1 .0 8 1 .1 0 1 .1 0

I-7 5 S B : S R I 8 2 1 to S R 8 2 6 4 .8 1 .4 2 1 .1 5 1 .0 8 1 .3 0 6 .6 5 2 .9 1 1 .2 9 4 .7 3

S R 8 2 6 E B : I-7 5 to I-9 5 8 .3 1 .3 5 1 .1 5 1 .2 1 1 .3 1 2 .9 4 1 .5 8 2 .7 4 2 .8 8

S R 8 2 6 W B : I-9 5 to I-7 5 8 .3 1 .4 0 1 .2 3 1 .3 2 1 .3 8 3 .4 2 2 .2 8 3 .6 4 3 .4 6

P l a n n i n g T i m e In d e xT r a v e l T i m e In d e x

8

7

6

5

4

3

2

1

15

highways in Florida during the PM peak period. The FAF data on AADTT for the same

roadways are displayed using variations in the thickness of the highway segments. Such analyses

can be conducted to identify and prioritize the locations with combinations of greater congestion

and truck travel demand.

Figure 2.8 Fusion of ATRI’s Freight Performance Measures FAF Data on Truck Volumes

to Identify Important Bottlenecks for Freight Movement

As another illustration, Figure 2.9 shows an example of a Florida urban interchange that

ranked in the top 100 freight bottlenecks in the nation. In this study (ATRI, 2009), each

bottleneck was given a “freight congestion index value”2 that was calculated based on congestion

at the location (measured by the extent to which the measured truck speeds are below free-flow

speeds) and the freight demand at the location (measured by the hourly freight truck volume).

Subsequently, the different locations were rank-ordered based on the congestion index value.

Such analysis can be conducted for Florida to identify and rank-order the highway bottlenecks in

the state for prioritization of funding toward fixing highway freight bottlenecks in the state.

2 A complete description of the methodology for identifying and rank-ordering the bottlenecks is available at

http://www.atri-online.org/fpm/ResearchMethodology.pdf, accessed on 3-10-2014.

http://www.atri-online.org/fpm/ResearchMethodology.pdf

16

Furthermore, one can analyze the specific segments of the bottleneck to identify which segments

or directions of an interchange, for instance, have the worst congestion.

Figure 2.9 Bottleneck Analysis for I-4 at I-275 in Tampa

Source: ATRI (2009) Bottleneck Analysis of 100 Freight Significant Highway Locations.

Available at: http://atri-online.org/2010/05/08/944/, accessed on 3-10-2014.

http://atri-online.org/2010/05/08/944/

17

CHAPTER 3 : CONVERSION OF ATRI TRUCK GPS DATA TO TRUCK TRIPS

3.1 Introduction

As discussed in the previous chapter, ATRI data comprised large streams of GPS records of

truck movements in North America. The raw GPS data streams from ATRI needed to be

converted into a truck trip format to realize the full potential of the data for freight flow analysis,

modeling, and planning applications. Development of such a truck trip database involved the

determination of truck starting and ending instances and locations, trip distance, total trip

duration, and duration of intermediate stops (e.g., at traffic signals and rest stops) in the trip.

Doing so required separation of valid pickup/delivery stops from congestion stops, stops at

traffic signals, and stops to meet hours of service regulations,3 making use of land-use

information and GIS analysis tools along with carefully-considered assumptions. In addition, the

process involved resolution of potential anomalies in GPS data, such as data discontinuities due

to loss of satellite signals. This chapter describes the algorithms and procedures developed in the

project to convert the raw GPS streams provided by ATRI data into a database of truck trips. The

next section provides a brief description of ATRI’s truck GPS data used in the project. Section

3.3 describes the algorithms and procedures. Section 3.4 presents the results from the algorithms.

3.2 ATRI’s Truck GPS Data

In this project, ATRI provided more than 145 million raw GPS records of truck movements from

four months—March, April, May, and June—in 2010 for the state of Florida. Specifically, for

each of these four months, all trucks from ATRI’s database that were in Florida at any time

during the month were extracted. Subsequently, the GPS records of those trucks were extracted

for time periods ranging from two weeks to an entire month as they traveled within Florida as

well as in other parts of North America. This allows the examination of truck movements within

Florida as well as truck flows into (and out of) Florida from (to) other locations in the nation.

Each GPS record contained information on its spatial and temporal location along with a

unique truck ID that did not change across all the GPS records of the truck for a certain time

period varying from one day to over one month (at least two weeks for most trucks in the data).

In addition to this information, a portion of the GPS records database contained spot speeds

information (i.e., the instantaneous speed of the truck for each GPS record), and the remaining

portion of the database did not contain spot speed information. In the remainder of this report,

the former type of data is called data with spot speeds and the latter type is called data without

spot speeds. The data with spot speeds and the data without spot speeds were separately

delivered to USF, presumably because they come from different fleets of trucks based on the

type of GPS units/technology used to monitor the trucks. The frequency (i.e., ping rate) of the

GPS data streams varied considerably, ranging from a few seconds to over an hour of interval

between consecutive GPS records. Table 3.1 shows the distribution of the time gap between

consecutive GPS readings in one week of data during the month of May 2010. Whereas a large

3 Long-haul drivers make en-route stops for long durations to rest before resuming further driving. Hours-of-service

regulations in 2010 by the Federal Motor Carrier Safety Administration (FMCSA) required an 11-hour maximum

daily driving limit. Specifically, truck drivers were allowed to drive up to 11 consecutive hours only after10 hours of

off-duty time (or rest). Therefore, not all stops of longer duration are valid pickup/delivery stops. See

http://www.fmcsa.dot.gov/regulations/hours-of-service for more information on FMCSA’s hours-of-service

regulations for different years.

http://www.fmcsa.dot.gov/regulations/hours-of-service

18

proportion (79.7%) of data with spot speeds comprised GPS streams at less than 15-minute

interval, a considerable proportion (29.5%) of data without spot speeds had GPS streams at

greater than 1-hour intervals.

Table 3.1 Distribution of Time Gap between GPS Readings

in a Week of Truck GPS Data % of Consecutive

GPS Readings in Data

Data with

Spot Speeds

Data without

Spot Speeds

< 1 minute 12.4% 21.5%

1–5 minutes 25.5% 15.0%

5–10 minutes 17.8% 8.5%

10–15 minutes 24.0% 4.9%

15–30 minutes 12.4% 9.7%

30 minutes–1 hour 2.8% 10.9%

1–2 hours 1.7% 27.8%

> 2 hours 3.4% 1.6%

For each GPS record, ATRI extracted and provided to USF information on how far the

location is from the nearest interstate highway. In addition, ATRI shared with USF a geographic

file (i.e., GIS shape file) containing polygons of major truck stops (such as rest stop areas, weigh

stations, welcome centers, and wayside parking) within and outside Florida.

3.3 Algorithm Description

The overall procedure to convert ATRI’s truck GPS data into a database of truck trips can be

described in the following five broad steps. Each of the broad steps is detailed in this section.

1) Clean, read and sort the GPS data for each truck ID into a time series, in the order of

the date and time of the GPS records.

2) Identify stops (i.e., trip-ends or trip origins and destinations) based on spatial

movement, time gap, and speed between consecutive GPS points.

a) Derive a preliminary set of trips based on a minimum stop dwell-time buffer

value (i.e., eliminate stops of duration less than dwell-time buffer value). Use

30 minutes dwell-time buffer in the beginning.

b) Combine very small trips (< 1 mile trip length) with preceding trips or

eliminate them.

c) Eliminate poor-quality trips based on data quality issues such as large time

gaps between GPS records and incomplete trips (i.e., trips spanning beyond

the temporal limits of the study period).

3) Eliminate trip-ends in rest areas and other locations that are unlikely to involve a

valid pickup/delivery.

a) Overlay trip ends on a geographic file of rest areas, wayside stops, and similar

locations.

b) Eliminate stops in close proximity of interstate highways, which are most

likely to be rest areas or wayside parking stops.

19

c) Join consecutive trips ending and beginning at such stops.

4) Find circular (i.e., circuitous) trips based on the ratio between air distance to roadway

network distance. Use raw GPS data between the origin and destination of circular

trips to split them into appropriate number of shorter, non-circular trips by allowing

smaller dwell-time buffers at the destinations. To do this, implement step 2 with a

smaller dwell-time buffer (15 minutes) and go through steps 3 and 4 to find any

remaining circular trips. Repeat the process with a dwell-time buffer of 5 minutes to

split remaining circular trips.

5) Conduct additional quality checks and eliminate trips that do not satisfy quality

criteria.

3.3.1 Clean and Sort Data

The raw GPS data were first screened for basic quality checks such as the presence of spatial and

temporal information and the presence of at least one day of data for each truck ID. Truck IDs

that did not have GPS data for at least a span of one day or that had too few GPS records were

removed. For such trucks, it was difficult to extract trips because most of the data were likely to

be lost in the form of incomplete trips, i.e., trips without a valid origin and/or destination in the

data. The cleaned data were then sorted in a time series for each truck, beginning from the GPS

record with the earliest date and time stamp.

3.3.2 Identify Truck Stops (i.e., Truck Trip-ends) to Generate Truck Trips

This step comprised a major part of the procedure to convert raw GPS data into truck trips. The

high-level details of the algorithmic procedure in this step are presented in Figure 3.1. Following

is a list of the terms used in the algorithm along with their definitions:

1) Travel distance (td): Spatial (geodetic) distance between two consecutive GPS

records.

2) Travel time (trt): Time gap between the two consecutive GPS records.

3) Average travel speed (trs): Average travel speed between consecutive GPS records

(td/trt).

4) Trip length (tl): Total distance traveled by the truck from origin of the trip to the

current GPS point. This becomes equal to trip distance, when the destination is

reached.

5) Trip time (tpt): Total time taken to travel from origin of the trip to the current GPS

point.

6) Trip speed (tps): Average speed of the trip between the origin and the current GPS

point.

7) Origin dwell-time (odwt): Total time duration of stop at the origin; i.e., when the

truck is not moving (the wait time for the truck before starting its trip)

8) Destination dwell-time (ddwt): Total duration the truck stops at the destination of a

trip.

9) Stop dwell-time (sdwt) – Duration of an intermediate stop (e.g., traffic stop).

10) Total stop dwell-time (tsdwt): Total duration at all intermediate stops during the trip.

20

For current truck ID, get first GPS record from database.

Set GPS record number x = 1

Initialize td, trt, trs, tl, tpt, tps, odwt, ddwt, sdwt,

tsdwt

Get next GPS record (x = x+1) . Compute td, trt, trs between x & x-1

Start

trs < 5 mph?or spot speed of

x-1 =0?

Y

Truck is moving. So update trip attributes.tl = tl+td; tpt = tpt+trt;

sdwt = 0

N

tl > 0 ?(has the truck

moved from the origin)

sdwt = sdwt + trt;tsdwt = tsdwt+trt

N

Truck at the trip origin. Update origin dwell time

(odwt = odwt + trt). Mark GPS record x-1 as

origin

Stop dwell time (sdwt + trt) > minimum dwell time buffer?

Y

N

Mark x-1 as destination for the trip. ddwt = ddwt+sdwt+trt; sdwt = 0

Y

tl > 1 mile ?(is this a significant trip?)

A previous trip exists for the truck

and the destination of that trip and origin of this trip are less than I mile

apart?

N

NDo not record this as

a separate trip

Add total time of this trip to dwell time of previous trip.

odwt+tpt+trt+ddwt = ddwt of previous trip

YRecord the trip, its origin, destination, start and end times, tl, tpt, and tsdwt.

Y

Truck still at the destination. Update destination dwell time

ddwt = ddwt+trtY

Truck started moving.Mark x-1 as origin for next trip. Initialize attributes for next trip.

tl = td; tpt = trt; odwt = 0; ddwt = 0; sdwt = 0; tsdwt=0

Discard insignificant trip. Mark x as origin for next

trip.

trs < 5 mph?or spot speed of

x-1 =0?

N

Get next GPS record (x = x+1) . Compute td, trt, trs between x & x-1

Figure 3.1 Algorithm for Identifying Truck Trip Ends from Raw GPS Data

21

The first three terms—td, trt, and trs—are measures of movement between consecutive

GPS data points. The next three terms—tl, tpt, and tps—are measures of total travel between the

trip origin and the current GPS data point. When the truck destination is reached, these measures

are for the entire trip beginning from its origin to the destination. The last four terms—odwt,

ddwt, sdwt, and tsdwt—are dwell-times (i.e., stop durations) at different stages during the trip.

odwt is the dwell-time at the origin of a trip, ddwt is the dwell-time at the destination of the trip,

sdwt is the dwell-time at an intermediate stop (e.g., traffic stop) that is not the destination of the

trip, and tsdwt is the sum of dwell-time at all intermediate stops during the trip.

For each truck ID, the algorithm begins with reading its first GPS record and initializing

all the terms—td, trt, trs, tl, tpt, tps, odwt, ddwt, sdwt, and tswdt. Then, the algorithm reads the

next record and computes average travel speed between the two records to verify if the truck is

moving or if it is at rest. The subsequent parts of the algorithm are described below.

3.3.2.1 Determining Truck Stops and Moving Instances

An important component of the algorithm involved determining whether a truck was at a stop

(i.e., rest) or in motion. As can be observed from the flowchart (Figure 3.1), the primary

condition used to determine whether a truck was at a stop (which could be an origin, a

destination, or simply an intermediate stop) or it was moving was based on the average travel

speed between consecutive GPS data points. A cut-off speed of 5 mph was used; if the average

travel speed between consecutive GPS records was less than 5 mph (i.e., if trs < 5 mph), then the

truck was assumed to have stopped. As mentioned earlier, a portion of the data contained spot

speeds (i.e., instantaneous speeds), and remainder of the data did not contain information on spot

speeds. We used the data with spot speeds to test different cut-off values on the travel speed

between two consecutive GPS data points. Besides, other recent studies that converted ATRI’s

truck GPS data into trips (Bernardin et al., 2011; Kuppam et al., 2014) also used a 5-mph cut-off

to determine whether a truck is moving or if it is at rest.

For data with spot speeds, the average travel speed criterion (i.e., if trs < 5 mph) was used

and was checked if the spot speed was zero. If one of these two criteria was satisfied, the truck

was assumed to be at rest. It was important to check for small travel speeds between consecutive

GPS data points (i.e., if trs < 5 mph), even in data with spot speeds. When the time gap between

two consecutive GPS data points that are moving (i.e., spot speed > 0) was large enough, the

only way to determine if the truck stopped between two points was by checking its travel speed

between the two points. Limited tests on data with spot speeds suggested that it was sufficient to

use the average travel speed criterion (i.e., without checking for spot speeds), increasing

confidence in the quality of trips derived from the data without spot speeds.

An alternative approach to determine truck stops was to use a minimum distance criterion

(i.e., assuming the truck to be at a stop if it did not move beyond a certain distance, i.e., distance

cut-off value). However, since the time gap between consecutive GPS records varied

considerably—ranging from a second to several hours—it was difficult to use a single distance

cut-off for determining truck stops.

22

3.3.2.2 Separating Intermediate Stops from Trip Destinations

Checking for average travel speeds between consecutive GPS records helped in identifying if the

truck was at rest or moving. However, truck being at rest did not necessarily indicate whether it

was at a valid origin or destination or if it was at an intermediate stop. The intermediate stops

could be due to a number of reasons, including stops at traffic signals and traffic congestion

(these stops typically tend to be of a few minutes duration), stops at gas stations for refueling

purposes, wayside stops for drivers’ quick relaxation and other purposes such as restroom visits,

and rest stops of long duration to comply with hours of service.

To identify and eliminate many such intermediate stops, once the algorithm detected a

stop (i.e., trs < 5 mph or spot speed =0) for a trip in progress (i.e., tl > 0), it started accruing the

stop dwell-time (sdwt) based on the time elapsed between successive GPS data points (sdwt =

sdwt + trt). If the truck started moving again (i.e., if trs > 5 mph or spot speed > 0) before the

stop dwell-time reached the minimum dwell-time buffer value, then the stop was considered an

intermediate stop, and the algorithm proceeded to find another stop. On the other hand, if the

stop dwell-time exceeded the minimum dwell-time buffer value, then the stop was considered a

candidate for valid destination, and the stop dwell-time (sdwt) was considered as part of the

destination dwell-time (ddwt), by updating ddwt as sdwt + trt. Subsequently, if the length of this

trip (tl) was greater than 1 mile (if not it was considered an insignificant trip and not recorded),

the trip was recorded, along with its origin, destination, start/end times, trip length, and the total

time the truck stopped at all intermediate locations between origin and destination (tsdwt).4 The

destination dwell-time of the trip was then updated using subsequent GPS records (i.e., ddwt =

ddwt + trt) until the truck started moving again (i.e., trs > 5 mph or spot speed > 0). Once the

truck started moving (see the bottom left portion of the flowchart in Figure 3.1), the origin of a

potential next trip was marked, and the algorithm proceeded to find the next truck stop, as can be

observed from the looping of the bottom portion of the algorithm to the top portion (on the left

side of Figure 3.1).

A major determinant of the quality of outputs from the above procedure depended on the

dwell-time buffer (i.e., the minimum stop duration required for a truck stop to be called a

destination). Some previous studies (Ma et al., 2011) considered stops of less than 3 minutes’

duration as intermediate stops while other studies (Bernardin et al., 2011; Kuppam et al., 2014)

considered stops of less than 5 minutes’ duration as intermediate stops. In this study, however,

based on discussions with ATRI and our own tests using Google Earth (i.e., by observing the

land-uses of stops made by trucks and the duration of those stops), larger dwell-time buffers

were used. This was because the focus of this research was to extract truck OD flows over long-

haul distances of concern for Florida’s statewide freight model. While most intermediate stops in

urban areas tend to be of smaller duration (e.g., most traffic signal cycles tend to be of less than 3

minutes’ duration), not all stops of larger duration tend to be for commodity pickups and/or

deliveries. Intuition suggests that 5 minutes is not sufficient for picking up or delivering many

types of goods. Besides, refueling stops, stops at weighing stations, and quick relaxation or

restroom stops tend to be longer than 5 minutes. And, of course, truck stops for the purpose of

complying with hours of service regulations tend to be of several hours duration.

4 Most previous studies do not record the total stop dwell-time (tsdwt) — a useful attribute for truck trips, especially

for understanding how the duration of intermediate stops varies with trip length and other relevant factors.

23

To arrive at an appropriate dwell-time buffer, we compared the trip outputs using

different values of dwell-time buffer—5, 10, 15, 30, and 45 minutes and 1 hour—with the land-

uses of the trip ends in Google Earth. Dwell-time buffers of short durations such as 5 minutes or

10 minutes were resulting in false identification of too many intermediate stops as truck origins

and/or destinations, while dwell-time buffers of too long durations led to missing valid

origins/destinations at pickup/delivery locations such as distribution centers. Further, using

smaller dwell-time buffers was leading to a larger share of short-length trips (because a long trip

between an origin and destination was broken down into several short trips). Besides, for low-

frequency data where the consecutive GPS records have large time gaps, small dwell-time

buffers cannot be relevant (for example, testing for a 5-minute dwell time would not give

different results than testing for a 15-minute dwell-time buffer if the raw GPS records are spaced

at 15-minute interval). After testing for different values of dwell-time buffers, we realized that no

dwell-time buffer value was perfect, and a trade-off had to be made between minimizing false

identification of unnecessary intermediate stops as trip ends, on one hand, and skipping of valid

origins and destinations on the other. After extensive tests via following trucks on Google Earth,

a 30-minute dwell-time buffer was used as a beginning point to separate intermediate stops from

trip destinations.

The 30-minute dwell-time buffer helped in avoiding most intermediate stops, including

traffic and congestion stops, wayside stops, gas refueling stops, and short stops at rest areas. But

it would not help eliminate longer duration stops at rest areas, including those made to comply

for hours-of-service regulations. Another issue was that the 30-minute dwell-time buffer led to

skipping of some valid origins or destinations that involved smaller dwell-times. These two

issues are addressed latter.

3.3.2.3 Dealing with Insignificant Trips

Trips that were too short in length were not recorded as independent trips. The minimum

acceptable trip length was assumed to be one mile. Therefore, if a trip was of length less than one

mile, it was discarded unless the trip occurred in the same area of the previous trip’s destination.

In this case, the insignificant trip’s time was simply added to the previous trip’s destination dwell

time (see bottom right portion of the algorithm in Figure 3.1). For example, if the destination of a

trip is large in size (such as a port), it might happen that the truck moves within the port for less

than one mile, leading to insignificant trips. Such movement was not considered a new trip but,

since the truck would still be at the same destination as the previous trip (port), it was

incorporated into the previous trip’s destination dwell time.

3.3.2.4 Quality Control Checks in the Algorithm

Figure 3.1 does not present all details of the algorithm to make it easier for readers to understand

the main components of the algorithm and for ease in presentation. These details include the

following quality checks embedded into the algorithm.

Dealing with large time gaps between consecutive GPS records: The ping rate in the data

(i.e., the frequency at which GPS positions are recorded) varied considerably, ranging from a few

seconds to several hours. Data with large time gaps between consecutive records could be due to

many reasons, including loss of GPS signals (e.g., in tunnels and mountains) and malfunctioning

24

of the GPS device. In such cases, the extracted trips tend to be of lower quality because it is

difficult to use only the spatial and temporal movement information to ascertain what happened

in the time gap. On the other hand, it is also possible that some GPS units (depending on the type

of equipment) may not record truck positions for an extended time period simply because the

truck engine is switched off. In such situations, the truck is simply not making any movements

for an extended time period. Therefore, if the time gap between two consecutive records was

greater than 2 hours and if the trip was in progress (i.e., the travel speed between the two records

was greater than 5 mph), such a trip was discarded. However if the speed was less than 5 mph,

then the truck was assumed to be at rest (i.e., not moving) for the entire time gap.

Trips spanning beyond the temporal limits of the study period: It is not necessary that the

GPS records of a truck begin with a trip origin and end with a trip destination. For many trucks,

the first several records indicated that the truck was in motion (because the trip started before the

first available GPS record for the truck) and/or the last records belonged to a trip that ended after

the last GPS record. Such incomplete trips found at the edges of study periods were discarded.

The above quality checks were implemented in the algorithm every time a GPS record

was read and the travel distance (td), travel time (trt), and average travel speed (trs) were

computed between consecutive records. In addition to the above quality controls embedded

within the algorithm, quality checks were conducted on the trips output at the end of the

procedure. Specifically, trip speeds, trip time, and distance were examined for manifestations of

any anomalies such as GPS jumps and jiggles. GPS jumps happen when GPS records show

unrealistically large movements within short durations, which manifest as trips with

unrealistically large speeds. Such trips were eliminated based on average trip speeds and travel

time. Specifically, only those trips within an average speed of 80 mph (between origin and

destination) were retained. Further trips that were too short in time (i.e., trips of travel time less

than a minute) also were removed.

3.3.3 Eliminate Trip Ends in Rest Areas

The above-described algorithm eliminates unwanted trip ends (such as traffic stops and refueling

stops) to some extent. However, the algorithm does not eliminate unwanted trip ends in rest areas

and other locations (e.g., wayside stops) with dwell times larger than the dwell-time buffer used

in the algorithm. To address this issue, the trip ends derived from the above step were overlaid

on a geographic file of rest stops provided by ATRI containing polygons of rest areas,

commercial truck stops, weigh stations, wayside parking, etc., throughout the nation. All the trip

ends falling in these polygons were eliminated by joining consecutive trips ending and beginning

in those polygons.

Doing the above helped in eliminating a large number of unwanted trip ends in rest areas,

wayside parking areas, and other such locations. However, further scrutiny suggested that a good

portion of trip ends were still in rest areas and similar locations. This is because the data in the

geographic file of rest stops provided by ATRI were not necessarily exhaustive of all rest areas

and other such stops (not for pickup/delivery) in the nation. To eliminate the remaining trip ends

in rest areas, the research team used information on the distance of each trip end from the nearest

interstate highway. When a random sample of 200 rest areas from the shape file provided by

ATRI were examined vis-à-vis their distance from the highway network, a vast majority of the

rest areas were found in very close proximity to interstate highways (45% were within 800 feet

25

distance of an interstate highway). Therefore, all trip ends within a buffer of 800 feet from

interstate highways were treated as stops at rest areas or wayside parking areas. Any consecutive

trips ending and beginning in the same location were joined to form a single trip.

The natural next question was how the 800-feet value was determined. We examined if

treating all trip ends within a close proximity of interstate highways as rest stops helps in

removing additional unwanted stops while not removing true origin or destination stops. To

examine this, the movement of 40 trucks was traced for at least two weeks on Google Earth and

observed the land-uses of their stop locations. For each of these 40 trucks, the number of valid

trip ends noticed in Google Earth were recorded and then compared with the number of trips

output from the algorithm after eliminating trip ends that were in the rest stops polygons of the

shape file provided by ATRI and those that were within a given proximity of interstate highways.

This was tested for different buffers around interstate highways—half mile, quarter mile (1320

feet), 1000 feet, 800 feet, and 500 feet. As expected, no single buffer was ideal; using a large

buffer led to elimination of too many valid origins/destinations, and using a small buffer led to

the presence of too many invalid origins/destinations such as rest areas or wayside parking areas.

However, using 800 feet provided a good trade-off between losing valid origins/destinations and

counting invalid origins/destinations.

3.3.4 Find Circular Trips and Split Them into Shorter Valid Trips

Recall from Section 3.3.2.2 that a 30-minute dwell-time buffer was used to separate intermediate

stops from valid trip destinations. Among the other dwell-time buffers examined, the 30-minute

buffer struck a good balance between removing intermediate stops (such as traffic signal stops,

congestion stops, and refueling stops) and skipping valid destinations (of less than 30 minutes

dwell time) en-route. However, it would be useful to recover such valid destinations of less than

30 minutes.

When the trip outputs from the algorithm with a 30-minute dwell-time buffer were

examined, some trips had origins and destinations that were too close to each other, although the

roadway network distance of those trips measured in the algorithm (using GPS data) was large.

Some trips, for example, had their origin and destination in the same location, although the

distance traveled by the truck on the roadway network was large. One reason for this was

because the algorithm was skipping some valid destinations (of less than 30 minutes dwell time)

en-route (and if the trucks were changing the travel direction after the skipped destinations).

One way to identify if a trip extracted from the algorithm had any valid destinations en-

route that have been skipped was to check its circuity ratio, the ratio between the air distance

between origin and destination and the roadway network distance between origin and

destination. The value of the circuity ratio can range from 0 to 1. If a truck travels in a straight

line between its origin and destination, its circuity ratio would be 1. However, most trips on the

highway transportation network tend to travel more than the air distance between the origin and

destination. At the same time, if the ratio is too small, there is a high chance that the truck

stopped en-route at valid destinations, albeit for shorter durations than 30 minutes. Intuitively, it

is unlikely that trucks detour significantly between the origin and destination only for the sake of

traveling to intermediate rest stops. Figure 3.2 shows an example of such a trip whose origin and

destination locations (marked by blue circles) are too close to each other, although the distance

26

traveled by the truck on the network along the route shown in red in the figure is more than 100

miles. In this example, the truck stopped at three other locations between the origin and

destination (marked by yellow circles) for less than 30 minutes. The land-uses of these stops,

when examined on Google Earth, were all valid destinations such as warehouses and large

grocery stores. When several other such examples were examined, it became more apparent that

trips with a small circuity ratio had skipped en-route stops of duration smaller than 30 minutes,

and most of these stops were valid destinations. After extensive testing, through tracing raw GPS

data of trips with different circuity ratio values, a cut-off value of 0.7 was determined. All trips

extracted from the algorithm with a circuity ratio less than 0.7 were considered to be circular

trips with a high chance of a skipped valid destination en-route.

The circular trips were then separated for further processing. Specifically, the procedure

went back to the raw GPS data of the trips with a circuity ratio less than 0.7 and re-applied the

algorithm in Figure 3.1 with a smaller dwell-time buffer (15 minutes). This helped in splitting

the circular trips into multiple potentially-valid trips. Specifically, each circular trip was broken

into an appropriate number of shorter, non-circular trips by allowing smaller stop dwell-time

buffers at the destinations. The trip outputs from this process were, again, checked for circuity.

For any remaining trips with a circuity ratio of less than 0.7, the algorithm in Figure 3.1 was

reapplied on the corresponding raw data, albeit with a smaller dwell-time buffer (5 minutes). In

the example in Figure 3.2, this iterative process would result in four separate trips, each with a

circuity ratio greater than 0.7, instead of a single trip with a small circuity ratio.

Figure 3.2 Example of a Circular Trip Extracted from the Algorithm

with 30-Minute Dwell-time Buffer

An example of the results from this procedure is in order here. Applying the algorithm in

Figure 3.1 with a dwell-time buffer of 30 minutes and the subsequent step (eliminating trip ends

27

in rest areas and in close proximity of interstate highways) to one month of ATRI’s GPS data

(May 2010) resulted in a total of 252,000 trips. Of all these trips, 183,000 (72.6%) had circuity

ratios greater than or equal to 0.7. After splitting the remaining 69,000 trips (with circuity ratio <

0.7) into smaller trips by reapplying the algorithm in Figure 3.1 with a minimum dwell-time

buffer of 15 minutes, about 123,000 trips were extracted. About half of these trips had a circuity

ratio of at least 0.7. For the other half, the algorithm in Figure 3.1 was repeated with a minimum

dwell-time buffer value of 5 minutes. This resulted in more than 87,000 trips, of which 38,000

trips had a circuity ratio of at least 0.7. The remaining 49,000 trips, with a circuity ratio smaller

than 0.7, were discarded. In all, the final number of trips extracted for the month of May 2010

was 183,000 + 62,000 + 38,000 = 283,000.

The above-described iterative procedure of checking for circuity and splitting circular

trips into multiple valid trips using smaller dwell-time buffers helped in two ways: (a) it helped

capture trips with valid destinations of short dwell times, and (b) it helped remove any remaining

circular trips, some of which were likely to have resulted from joining consecutive trips within a

close proximity of interstate highways, as described in the previous section.

3.4 Results

The above-discussed procedure was applied to four months of raw data (March–June 2010)

comprising more than 145 million raw GPS records. This resulted in more than 2.7 million truck

trips. Of these, more than 1.27 million trips had at least one end in Florida. Table 3.2 shows a

summary of the raw data and the trips derived from the data. Summaries are provided for each

month of the data, separately for data with spot speeds and data without spot speeds. The number

of trips extracted, the number of unique truck IDs to which these trips belonged, and the average

trip distance and trip speeds (without considering duration at rest stops) are presented for three

different types of trips—(a) all trips including those outside Florida, (b) FL-link trips (trips with

at least one end in Florida), and (c) FL-only trips (trips with both origin and destination in

Florida).

Note from the table that the trips extracted from data without spot speeds were longer

than those from data with spot speeds. For a certain type of data (e.g., for data with spot speeds),

the average trip distances and trip speeds were similar across the four months. Besides, the

average trip speeds appeared to be similar across different datasets and for different months. A

detailed analysis of the characteristics of the trips extracted in this project is presented in the next

chapter.

The trip outputs from the procedures discussed in this chapter were subject to a variety of

quality checks, some of which are discussed here. The land uses of the OD locations of a random

sample of 232 trips extracted from the algorithms were examined on Google Earth. More than 90

percent of the 464 trip ends were in locations that are highly likely to involve goods

pickups/deliveries, such as distribution centers, manufacturing companies, industrial areas, ports,

retail stores, shopping centers, and agricultural lands. Of the remaining locations, 24 were on

highways (that are not interstate highways) without nearby freight-related land uses, 3 were in

rest areas, and 9 were in gas stations. Most of the 24 trip ends on highways were truck stops of

longer than 30 minutes. In future work, eliminating truck stops in close proximity to major

highways (in addition to interstate highways) potentially can improve the results. For stops at gas

28

stations, however, particularly those of greater than 30 minutes’ duration, it is difficult to

decipher if they are made for refueling purposes or for fuel delivery services. In addition to trip

end locations, the accuracy of temporal attributes (i.e., trip start and end times) was assessed. The

trip start times output from the algorithm were found to be accurate for more than 95 percent of

the trips, and the trip end times were accurate for all trips. Overall, while scope exists for

improving the algorithms in this chapter (e.g., by using detailed land-use information), the

quality of trips extracted suffices for the purpose of estimating statewide TAZ-to-TAZ truck

flows.

29

Table 3.2 Summary of Truck Trips Extracted from Four Months of ATRI’s Truck GPS Data

Data with Spot Speeds Data without Spot Speeds All Data

All Trips FL-link

Trips

FL-only

Trips All Trips

FL-link

Trips

FL-only

Trips All Trips

FL-link

Trips

FL-only

Trips

March

2010

Number of GPS records 13,271,519 25,750,534 39,022,053

Number of trips extracted 284,092 145,245 119,602 449,074 195,298 128,178 733,166 340,543 247,780

Number of unique truck IDs 7,406 6,594 4,815 47,523 39,277 25,979 54,929 45,871 30,794

Average trip length (miles) 188 135 59 258 225 78 231 187 69

Average trip time (minutes) 212 162 83 315 286 120 275 233 102

Average trip speed (mph) 41 37 34 40 38 33 40 38 33

April

2010







May

2010







June

2010







All

Four

Months


Number of trips extracted 1,144,048 584,612 478,199 1,563,633 687,234 450,486 2,707,681 1,271,846 928,685





30

CHAPTER 4 : CHARACTERISTICS OF TRUCK TRIPS

DERIVED FROM ATRI DATA

4.1 Introduction

This chapter presents an analysis of the truck trip data derived from the four months of ATRI’s

truck GPS data described earlier. The truck travel characteristics analyzed included trip duration,

trip length, trip speed, time-of-day profiles, and origin-destination flows. Each of these

characteristics was derived at a statewide level and for different regions in the state—

Jacksonville, Tampa Bay, Orlando, Miami, and rest of Florida—defined based on the freight

analysis framework (FAF) zoning system. Several of these characteristics are provided in

Appendix B. Furthermore, the chapter presents an analysis of Origin-to-Destination (OD) travel

distances, travel times, and travel routes between a selected set of TAZ-to-TAZ OD pairs.

Appendix C augments this chapter by providing route choice and travel time distributions

derived from the data for 10 different OD pairs.

4.2 Trip Length, Trip Time, Trip Speed, and Time-of-Day Distributions

Figure 4.1 shows the trip length distribution of more than 2.7 million trips derived from the data.

As can be observed, a considerable proportion of the trips are within 50 miles’ length. Figure 4.2

shows the trip duration distribution of these trips. Two types of trip durations are reported: (1)

total trip time and (2) trip time in motion. Total trip time is the time between trip start and trip

end, including the time spent at rest stops. Trip time in motion excludes the time spent at rest

stops and other long-duration stops. Note that trip time in motion includes time at smaller

duration (< 5 minutes) stops such as traffic stops to reflect congestion effects. Figure 4.3 shows

the trip speed distribution considering the two types of trip times discussed above. Specifically,

the average trip speed considers all stops between trip start and trip end, and trip speed in motion

excludes stops of longer duration (e.g., rest stops) but considers stops of smaller duration.

31

Figure 4.1 Trip Length Distribution of All Trips Derived from Four Months of ATRI Data

Figure 4.2 Trip Time Distribution of All Trips Derived from Four Months of ATRI Data

0

5

10

15

20

25

30

35

40

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

100

0-2

000

200

0-3

000

>3

00

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

5

10

15

20

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

30-4

20

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

32

Figure 4.3 Trip Speed Distribution of All Trips Derived from Four Months of ATRI Data

Appendix B provides the distributions of trip length, duration, speed, and time-of-day5

profiles for different segments of the 2.7 million trips discussed above. The different segments

include trips starting and ending in different FAF zones in Florida—Jacksonville FAF zone,

Tampa FAF zone, Orlando FAF zone, and Miami FAF zone. Following are the specific counties

in each of these FAF zones:

Jacksonville FAF zone: Baker, Clay, Duval, Nassau, St. Johns

Miami FAF zone: Broward, Miami-Dade, Palm Beach

Orlando FAF zone: Flagler, Lake, Orange, Sumter, Osceola, Seminole, Volusia

Tampa FAF zone: Hernando, Hillsborough, Pasco, Pinellas

The distributions are provided separately for weekday and weekend trips. Such

distributions potentially can be used for modeling heavy truck trip characteristics within the

major regional models in the state. In addition to the above distributions, for each urbanized

county in each of these FAF zones, the top 10 origins and destinations are provided at the state-

level and the county-level geography. As an example, the truck trip characteristics for the Tampa

FAF zone are provided in Figures 4.4 through 4.9, and those for other FAF zones are provided in

Appendix B. It is interesting to note that the time-of-day profiles for all the four FAF zones in

Florida showed a single peak during the late morning period as opposed to a bi-modal peak

typically observed for passenger travel for morning and evening peak periods.

5 Note: Time-of-day of a trip is determined based on the hour in which the midpoint of the trip falls.

0

5

10

15

20

25

30

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)

Trip Speed in Motion

Average Trip Speed

33

The appendix provides truck trip characteristics for the following trip segments as well:

(a) trips that start and end in Florida (Internal–Internal trips for Florida), (b) trips that start in

Florida but end outside Florida (Internal–External trips), and (c) trips that start outside Florida

and end in Florida (External–Internal trips). As an illustration, Figures 4.10 and 4.11 show the

top 10 destination states for trips starting from Florida and the top 10 origin states for trips

ending in Florida, respectively.

34

(a)

(b)

Figure 4.4 Trip Length Distribution of Trips Starting and Ending in Tampa FAF Zone

during (a) Weekdays (61,465 trips), and (b) Weekend (7,780 trips)

0

10

20

30

40

50

60

1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

35

(a)

(b)

Figure 4.5 Trip Time Distribution of Trips Starting and Ending in Tampa FAF Zone


0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

36

(a)

(b)

Figure 4.6 Trip Speed Distribution of Trips Starting and Ending in Tampa FAF Zone


0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (miles per hour)


Average Trip Speed

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps



Average Trip Speed

37

(a)

(b)

Figure 4.7 Time-of-Day Profile of Trips Starting and Ending in Tampa FAF Zone


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps

Time of Day (0 is 12 AM)

Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

38

Figure 4.8 Destinations of Trips Starting in Hillsborough County (87,701 Trips)

Top 5 destinations among

other states

Georgia

Alabama

North Carolina

South Carolina

Tennessee

Top 10 destinations other than

Hillsborough County

Polk

Pinellas

Orange

Manatee

Duval

Pasco

Lee

Sarasota

Lake

Miami-Dade

39

Figure 4.9 Origins of Trips Ending in Hillsborough County (86,254 Trips)

Top 10 origins other than Hillsborough

County

Polk

Pinellas

Orange

Pasco

Manatee

Duval

Lee

Sarasota

Osceola

Miami-Dade

Top 5 origins among other

states

Georgia

Alabama

North Carolina

South Carolina

Tennessee

40

Figure 4.10 Destinations of Trips Starting in Florida (1,102,873 Trips)


Florida

Georgia

Alabama

South Carolina

North Carolina

Mississippi

Texas

Tennessee

Louisiana

Virginia

Pennsylvania

41

Figure 4.11 Origins of Trips Ending in Florida (1,097,614 Trips)

Top 10 origins other than Florida

Georgia

Alabama

South Carolina

North Carolina

Texas

Tennessee

Mississippi

Louisiana

Pennsylvania

Virginia

42

4.3 Origin-Destination Truck Travel Time and Route Measurements

One of the tasks in the project involved the exploration of the use of ATRI’s truck GPS data to

generate information on truck travel time and routes between several locations (i.e., origin-

destination pairs) in the state. Such information can be used in future studies to analyze the route

choice behavior of trucks, to inform truckers of the time that can be expected to travel between

given OD pairs, and, potentially, to derive travel time reliability measures. In addition, truck

travel time skims between OD pairs can be generated for use in the Florida Statewide Model

(FLSWM).

The travel time skims currently used for the FLSWM are based on free-flow speeds. It

will be useful to update these travel time skims using the GPS data, because the GPS data

potentially can provide a better estimate of travel times. However, it is a cumbersome task to

estimate the travel times for each (and every) OD pair in the FLSWM. Given there are about

6,000 zones in the FLSWM, it would be 6000 × 6000 OD pairs (i.e., 36 million OD pairs). Thus,

it is not feasible to estimate travel times separately for each OD pair using GPS data. It is feasible

to measure the travel times for a smaller number of OD pairs. The estimated travel times have

been compared with the travel time skims currently used in the FLSWM and with the travel

times reported in Google Maps. Based on these comparisons, an assessment was made on the

currently used travel time skims in the FLSWM along with recommendations for future work on

estimating travel time skims for the FLSWM.

4.3.1 Factors Influencing Travel Distance and Travel Time Measurements

from ATRI’s GPS Data

A variety of factors influence measurement of travel distance and travel time measurements from

GPS data. These are discussed below.

Route Choice: Between any given OD pair with several truck trips extracted from the

ATRI data, it was observed that not all trips took the same route. Differences in the paths taken

by the trucks lead to differences in travel distances and travel times across different trips between

the same OD pair.

Ping rate of GPS data: Ping rate is the time gap between two consecutive GPS records.

The simplest method to measure OD travel distances using GPS data was by adding up distances

between every two consecutive GPS records, starting from the first GPS record at the origin and

ending with the last GPS record at the destination.6 The accuracy of the OD travel distances

measured using this method depends on how closely located are the consecutive GPS records (or

the ping rate of the GPS data). For GPS data with larger ping rates, the travel distance was

underestimated, since the air-field distance between consecutive GPS records was smaller than or

equal to the network distance. In ATRI’s truck GPS data, the data without spot speeds had much

higher ping rates (i.e., larger gap between consecutive GPS data points) than the data with spot

speeds. Therefore, only the data with spot speeds were used for measuring travel distances and

travel times and for deriving travel routes. Figure 4.12 shows the distribution of ping rates from

one week of ATRI data with spot speeds and without spot speeds (one distribution for each type

6 Better methods involve map-matching of the GPS coordinates to the roadway network and then measuring the

distance along the network.

43

of data). Within the data with spot speeds, only those trips with a maximum of 15 minute ping

rate were selected.

Figure 4.12 Distribution of Ping Rates from One Week of ATRI Data

with and without Spot Speeds (May 2010)

TAZ size: The size of TAZ at the origin and/or destination ends of the trip did not

actually influence the travel time measured from the GPS data, because the GPS data provided

more spatially detailed information (i.e., longitude and latitude) on trip ends. However, when

comparing the distances measured from the ATRI data with those from the distance

measurements in FLSWM, the TAZ size influenced the distance measured in FLSWM. This is

because in FLSWM the TAZ-to-TAZ distance is measured based on centroid-to-centroid

distance. In GPS data, not all trips end at the centroid. Therefore, depending on where the trips

end in a TAZ, the differences between the trip distances and times measured in GPS data to those

from the FLSWM may vary considerably. To minimize this variation, OD pairs that have at least

one end in Florida were selected (because the TAZs outside Florida are larger in size than those

inside Florida).

Time of day: Due to temporal variation of traffic congestion, the time-of-day during

which a trip is taken influences the travel time of the trip. In this study, the following time

periods were defined based on the midpoint of the trips:

0

5

10

15

20

25

30

35

40<

5

5-1

0

10

-15

15

-20

20

-25

25

-30

30

-45

45

-60

60

-90

90

-12

0

12

0-1

50

15

0-1

80

>1

80

Pe

rce

nta

ge

Ping Rate (minutes)

Data with spot speeds (One week in May 2010)

Data without spot speeds (One week in May 2010)

44

AM peak trips: trips with their midpoints between 6 a.m. and 9 a.m. (Florida time),

Mid-day trips: trips with their midpoints between 9 a.m. and 4 p.m. (Florida time),

PM peak trips: trips with their midpoints between 4 p.m. and 7 p.m. (Florida time),

and

Night-time trips: trips with their midpoints between 7 p.m. and 6 a.m. (Florida time).

Number of trips (extracted from GPS data) between OD pairs: For constructing

statistically-representative distributions of travel time between OD pairs, and for capturing the

variance in route choices and congestion effects, OD pairs with at least 20 trips extracted from

the ATRI data were selected for further analysis.

Table 4.1 shows the distribution of different OD pairs in FLSWM by the number of trips

extracted (from four months of ATRI data). A total of 327,326 OD pairs had at least one trip

extracted from the entire four months of ATRI data—with spot speeds and without spot speeds.

After eliminating the OD pairs with only trips from data without spot speeds, 126,138 OD pairs

had at least one trip extracted from data with spot speeds. Subsequently, eliminating OD pairs

with both ends outside Florida resulted in 75,089 OD pairs with at least one trip from data with

spot speeds (with at least one end in Florida). From these OD pairs, only 1,237 OD pairs with at

least 20 trips in the data were selected (shown in grey-shaded cells in the table). More than

60,000 trips were extracted from the data with spot speeds between these 1,237 OD pairs, with at

least one end in Florida and with at least 20 trips. For these trips, we were confident about the

measurement of travel distances and travel times and the extracted travel routes.

Table 4.1 Distribution of OD Pairs Based on Number of Trips

Extracted from ATRI Data

Number of Trips

from ATRI Data

# OD Pairs Based

on All Trips

from ATRI Data with &

without Spot Speeds

# OD Pairs Based on

Trips from ATRI Data

with Spot Speeds

All Trips from ATRI

Data with Spot Speeds

with at Least One

End in Florida

1–5 286,579 110,269 68,128

5–10 18,059 7,297 3,864

10–20 11,166 4,543 1,860

20–30 4,026 1,512 495

30–40 2,143 764 243

40–50 1,219 412 132

>=50 4,134 1,341 367

Total 327,326 126,138 75,089

4.3.2 Truck Travel Distance, Travel Time, and Travel Route Measurements

for 1,237 OD pairs

For the 1,237 OD pairs discussed above, the distributions of truck travel distance and travel times

were derived and summarized in an Excel file. The file consists of the following information for

each OD pair:

45

Origin and destination locations (FLSWM TAZ numbers, county, and state),

OD travel distance and free-flow travel time used as input to FLSWM, and

Distributions of OD travel distance, travel time in motion (i.e., travel time excluding

large duration stops at rest stops etc.), and corresponding travel speeds derived from

the ATRI data, expressed as minimum, average, and maximum values, along with 5th

,

10th

, 15th

, 50th

and 85th

percentile values.

In addition to travel distance and travel route measurements, the travel routes for a large

number of trips between these OD pairs were derived in the form of on-route network links

between the origin and destination TAZs. Specifically, the travel route information was extracted

for 36,703 trips between 725 OD pairs within Florida. For each of these trips, the travel route is

provided in the following fashion:

Origin TAZ,

Destination TAZ, and

Several GPS coordinates that fall on-route between the origin TAZ and the

destination TAZ, with each GPS coordinate representing the centroid of a network

link on-route between the origin TAZ and the destination TAZ7

The travel routes, travel distances, and travel times were examined in more detail for a

sample of 10 OD pairs. The 10 OD pairs were selected strategically to include a randomly

selected sample of 2 OD pairs in each of the following five travel distance bands: within 100

miles, 100 to 200 miles, 200-500 miles, 500-1000 miles, and above 1000 miles. For each of these

OD pairs, the route choice of all trips extracted from the data are depicted as on-route GPS

coordinates in Appendix C. Specifically, Figures C.1 through C.10 show maps with the route

choices, one figure for each OD pair. For illustration, the travel routes for 365 truck trips

extracted from the ATRI data between a sample OD pair are presented in Figure 4.13, in the

form of on-route GPS points between the origin and destination TAZs.

Figure 4.13 Route Choice for 365 Trips between a Sample OD Pair (3124-3703)

7 The FLSWM network was used for this purpose. Specifically, for each trip, each on-route GPS coordinate of the

trip was replaced by another GPS coordinate representing the centroid of the nearest network link.

Destination

TAZ#3703

Polk, FL

Origin

TAZ#3124

Orange, FL

46

One can observe from Figure 4.13 and other figures in Appendix C that a large proportion

(if not all) of trips between an OD pair tend to travel by largely similar routes. Even if the trips

may not travel by the exact same route, considerable overlap can be observed among the routes

across a large number of trips between an OD pair. Upon further examination of truck travel

routes for a few randomly-sampled OD pairs (not shown in figures), it can be seen that truck

travel routes tend to exhibit higher variability for travel over shorter distances within urban areas

than for travel over longer distances. These observations have interesting implications for future

research on understanding and modeling truck route choice. While this study did not delve

further into understanding the route choice patterns of long-haul trucks, this is an important area

for future research using the truck GPS data from ATRI.

For the same set of 10 OD pairs, the distributions of travel times (i.e., travel time in

motion including stops of smaller duration such as traffic stops and fueling stops but excluding

significant stops such as rest stops) and travel distances measured from ATRI’s truck GPS data

are shown in Table 4.2 along with the travel times and travel distances used as inputs for the

FLSWM and those measured in Google Maps. The measurement of travel distances and travel

times on Google Maps for each OD pair was done for the same route taken by the shortest trip

extracted from ATRI’s GPS data for that same OD pair. Several observations can be made from

the table, as discussed below.8

First, the travel distances (for the shortest trips) measured from the ATRI data were

smaller than those in the FLSWM and Google Maps. This was likely because the travel distances

computed from the GPS data were straight line approximations (and therefore, underestimations)

of the distances between consecutive GPS points. However, such underestimation of network

distances using the GPS data was smaller for shorter length trips. The reader will note that

underestimating distances should not influence the measurement of travel times using GPS data.

While the distance between consecutive GPS points was approximated (and thereby

underestimated) as a straight line distance, the time between consecutive GPS points was simply

the time gap between the consecutive time stamps (which depends on the actual route taken to

travel between the two coordinates).

8 To augment this table, Appendix C provides histograms of OD travel time distributions for the same 10 OD pairs.

Specifically, the distributions for three different types of travel times are provided: (1) total travel time, which

measures the time between trip start and tripe end including the time spent at all intermediate stops such as rest

stops, traffic congestion stops, and fueling stops; (2) travel time in motion ,including non-significant stops, which

measures the travel time excluding significant stops such as rest stops but includes non-significant stops, such as

traffic congestion stops and fueling stops; and (3) travel time in motion excluding non-significant stops, which

excludes the time spent at all intermediate stops including rest stops, traffic congestion stops, and fueling stops.

These distributions are provided for different time periods of the day such as AM peak, mid-day, PM peak, and night

time.

47

Table 4.2 Travel Time and Travel Distance Measurements for a Sample of 10 OD Pairs

OD pair

Travel Time (minutes) Travel Distance (miles)

FLSWM Google

Maps

ATRI data (Travel time in motion excluding traffic stops and

other small duration stops) FLSWM

Google

Maps

ATRI data

Shortest 15th

percentile

50th

percentile

85th

percentile Average Shortest

15th

percentile

50th

percentile

85th

percentile Average

3124-

3703 64 67 70 81 89 102 92 66.7 69.2 62.2 64.9 66.8 69 67.4

4526-

1863 54 67 62 65 74 94 80 55 56.4 55 57.2 57.8 59 59.3

2228-

4227 104 106 122 154 174 194 175 117.3 116 113.4 117.5 121 124.5 121.2

3662-

3124 117 123 137 144 154 167 156 130.1 133 124.5 131.5 133.4 136.6 133.8

5983-

792 209 224 231 252 260 278 267 247 244 240.2 243 248.5 250 249.1

2420-

4147 211 237 228 248 269 291 269 265 223 203 207 209.1 211.3 209.6

4035-

5819 497 553 631 645 683 723 689 652 654 649 650.1 653.4 675.5 660.1

5073-

6117 425 534 494 573 664 805 678 555 543 462 558 604.3 669.6 605.7

413-

6086 810 891 993 1063 1224 1347 1217 1130 1064 1120 1133 1164.1 1249 1192.6

2355-

6176 952 1159 1344 1361 1422 1475 1426 1338.8 1289 1335 1340.5 1354.3 1388 1359.3

48

Second, the shortest travel times measured using GPS data were considerably higher than

those used as inputs into the FLSWM or those measured in Google Maps. To examine this

further for a larger sample of OD pairs, the travel times measured using the data were plotted

against those measured from Google Maps and those used in FLSWM for 100 OD pairs. Figure

4.14 shows the plot comparing the travel times measured using GPS data with those derived

from Google Maps (on the same routes as in the GPS data and during the same time-of-day). In

this figure, the travel times measured using GPS data exclude the time spent at all en-route stops

including rest-stops and traffic signal stops. It can be observed that the truck travel times

measured using GPS data were higher than those from Google Earth (on the same travel routes).

The differences were higher for longer trips. These differences likely were because the travel

times reported in Google Maps are not necessarily exclusively for trucks; they are based on a

variety of different information sources, most of which are likely to be oriented toward passenger

cars. ATRI’s truck GPS data, on the other hand, were exclusively for trucks that tend to travel at

a slower speed than cars due to at least the following three reasons: (a) trucks are not allowed to

travel on the fastest lanes in many sections of the highway network, (b) trucks accelerate at a

slower rate than cars, and (c) truck drivers are often instructed (and incentivized) by the trucking

fleet owners to travel at moderate speeds for reducing fuel consumption and cost savings.

Therefore, it is reasonable that the truck travel times measured using the GPS data were larger

than the travel times measured by Google Maps.

Figure 4.14 Comparison of OD Travel Times Measured in ATRI Data

(excluding all stops en route) with Travel Times from Google Maps

Figure 4.15 shows a plot comparing the truck travel times measured using GPS data with

travel times used as inputs for the FLSWM (i.e., the travel times from the travel time skim

matrices). The truck travel times measured using GPS data were much higher than those in

0

200

400

600

800

1000

1200

1400

1600

0 200 400 600 800 1000 1200 1400 1600

OD

tra

vel

tim

e o

bta

ined

fro

m G

oo

gle

Ma

ps

for

the

Sa

me

Ro

ute

s (m

inu

tes)

OD Travel Time in Motion (excluding time spent at all on-route stops and at

traffic signal stops) from ATRI Data (minutes)

49

FLSWM. The differences increase with increase in the trip length. These results suggested that

the travel time skims used as inputs for the FLSWM, which are free-flow travel times for

passenger cars derived from speed limits on individual links on the shortest path between the

origin and destination, did not closely represent truck travel conditions on the network. In this

context, ATRI’s truck GPS data provides a significant opportunity to develop better information

on truck travel time skims used as inputs for the FLSWM. In this project, the truck travel time

information was derived for 1,237 OD pairs. Since there is a large number of OD pairs (more

than 25 million within Florida) in the FLWSM, it is not feasible to rely on the GPS data alone to

generate truck travel time information for all OD pairs. Instead, the travel time information

derived in this project for 1,237 OD pairs potentially can be used in future work to derive (or

impute) truck travel time information for all other OD pairs in the FLSWM. Such better truck

travel time information can be used to replace the currently used passenger car travel time skim

matrices used as inputs for the freight components of FLSWM.

Figure 4.15 Comparison of OD Travel Times Measured in ATRI Data

(excluding all stops en route) with Travel Times used in FLSWM

0

200

400

600

800

1000

1200

1400

1600

0 200 400 600 800 1000 1200 1400 1600

OD

Tra

vel

Tim

e U

sed

as

Inp

uts

fo

r th

e

FL

SW

M (

min

ute

s)

OD Travel Time in Motion (excluding time spent at all on-route stops and at

traffic signal stops) from ATRI Data (minutes)

50

CHAPTER 5 : ASSESSMENT OF THE COVERAGE OF TRUCK FLOWS IN FLORIDA

BY THE ATRI DATA

5.1 Introduction

ATRI’s truck GPS dataset served as the source data for this analysis. While not a census of all

truck movement within Florida, the substantial dataset proved valuable to understanding freight

movement within the state. ATRI’s truck GPS dataset, however, is not necessarily a random

population of the trucks in the state. Therefore, this chapter provides an assessment of the type of

trucks included in ATRI data and the geographic coverage of the data.

5.2 Truck Types in ATRI Data

Based on the discussions with ATRI and anecdotal evidence, it is known that the major sources

of ATRI data are large trucking fleets, which typically comprise tractor-trailer combinations.

According to the FHWA vehicle type classification, these are tractor-trailer trucks in class 8–13

categories.9 However, a close observation of the data, through following the trucks on Google

Earth and examining some travel characteristics of individual trucks, suggested that the data

included a small proportion of trucks that did not necessarily haul freight over long distances.

Since the data did not provide information on the vehicle classification of each individual truck,

some heuristics were developed in the project to classify the trucks into heavy trucks and

medium trucks. The heuristics, as explained below, are based on the travel characteristics of

individual trucks over extended time periods (i.e., at least two weeks).

As discussed earlier, more than 2.7 million truck trips were derived from about four

months of ATRI’s raw GPS data for the year 2010. This database included 169,714 unique truck

IDs. Since the raw GPS data for each truck were available for at least two weeks (up to one

month, in most cases), trucks that did not make at least one trip of 100-mile length in a two-week

period were removed from the data. In this step, 88,869 trips made by 7,018 unique truck IDs

were removed. The median length of such removed trips was 20 miles, suggesting the short-haul

nature of these trucks. Subsequently, trucks that made more than five trips per day were

removed, assuming that these trucks are not freight-carrying, tractor-trailer combination trucks.

In this step, 275,224 trips made by 918 unique truck IDs were removed. The median length of

these trips was 16 miles. For the reader’s information, a histogram of the distribution of the

trucks in the database by their daily trip rates is provided in Figure 5.1.

9 Most freight in the U.S. is carried by tractor-trailer trucks of five axles or more (i.e., class 9 or above) and some on

tractor-trailer units of less than five axles (i.e., class 8 trucks). See page ES-7 of

http://www.fhwa.dot.gov/policy/vius97.pdf .

http://www.fhwa.dot.gov/policy/vius97.pdf

51

Figure 5.1 Distribution of Trucks by Daily Trip Rate

After the above-discussed procedures, more than 2.34 million trips extracted from GPS

data of more than 161,776 unique truck IDs were considered as trips made by heavy trucks that

carry freight. These trips were further used for OD matrix estimation later in the project. The

trucks making these trips are considered to be long-haul trucks or heavy trucks (assumed to be

FHWA vehicle classes 8–13). The remaining trucks (7,936 truck IDs) whose trips were removed

from further consideration in OD matrix estimation were assumed to be short-haul trucks or

medium trucks that serve the purpose of local delivery and distribution.

It is worth noting here that the procedures used in this project to separate heavy trucks

from other trucks were simplistic. It is not necessary that only heavy trucks carry freight over

long distances while only medium trucks serve the purpose of short-distance delivery and

distribution services. Further research is needed to identify the composition of trucking fleet in

the ATRI data and the purposes served by those trucks in the data.

5.3 What Proportion of Heavy Truck Traffic Flows in Florida Is Captured

in ATRI’s Truck GPS Data?

ATRI’s truck GPS data represented a large sample of truck flows within, coming into, and going

out of Florida. However, the sample did not necessarily comprise the entire population of truck

flows. Also, it was unknown what proportion of truck flows in the state was represented by these

data. To address this question, truck traffic flows in one week of ATRI’s truck GPS data were

compared with truck counts data from Florida Telemetered Traffic Monitoring Sites (TTMS)

truck traffic counts for that same week. This section describes the procedure and results from this

analysis.

0

10

20

30

40

50

60

< 1

1-2

2-3

3-4

4-5

5-6

6-7

7-8

8-9

9-1

0

10

-11

11

-12

12

-13

13

-14

> 1

4

Pe

rce

nta

ge o

f Tr

uck

s

Number of Trips per Day

52

One week of ATRI’s truck GPS data, from May 9–15, 2010, were used to derive weekly

ATRI truck traffic volumes through the FDOT TTMS traffic locations. Specifically, all truck

GPS records available with ATRI for that week within Florida, as well as 60 miles beyond the

Florida border into Alabama and Georgia, were used. Including GPS data points 60 miles beyond

the Florida border helps account for truck trips coming into and going out of the state.

Generating data on weekly ATRI truck traffic volumes at each TTMS location required

counting the number of times the trucks in the ATRI data crossed the location in the week. To do

so, we first attempted to run the raw GPS data through map-matching algorithms embedded in

the network analyst tool of ArcGIS. However, given the sheer size of the raw GPS data, it was

practically infeasible to run all the raw GPS data through the ArcGIS map-matching tool.

To address the above issue, the raw GPS data records were reduced into a database of

truck trip ends and intermediate GPS data points at a 5-minute interval (using the earlier

discussed algorithms for converting the data into truck trips). Since the purpose was only to

reduce the data to a manageable size and not to derive true pickup/delivery trip origins and

destinations, any truck stop of more than 5 minutes dwell time was considered to be a trip end. A

truck was considered to be “stopped” if either the spot speed was zero for at least 5 minutes or

the average travel speed between consecutive GPS data points was less than 5 mph for at least 5

minutes. For each truck trip derived in this fashion, given its origin and destination location

coordinates, intermediate GPS data points in “motion” were sampled from the raw data at a time

interval of 5 minutes. An intermediate GPS data point was considered to be in motion if the truck

was moving at a speed of greater than 5 mph. In all, the raw GPS data was reduced into a

manageable GPS dataset comprising truck trip ends and intermediate GPS data points in motion

at 5-minute time interval.

For each truck trip derived in the above-described fashion, the trip ends and intermediate

GPS points were map-matched to the FLSWM highway network using the network analyst tool

in ArcGIS. The map-matching algorithm snaps the GPS points to the nearest roadway links and

also determines the shortest path between consecutive GPS data points. Since intermediate GPS

data points between the trip ends were sampled at only a 5-minute interval, this procedure results

in a sufficiently accurate route for the trip. The output from this process was an ArcGIS layer

containing the travel routes for all trips generated from ATRI’s one-week truck GPS data. This

ArcGIS layer was intersected with another layer of FLSWM network containing the TTMS

traffic counting stations (specified in the form of network links on which the TTMS stations

were located). This helped estimate the number of truck trips (or individual trip routes) crossing

each TTMS count station, which is nothing but the volume of ATRI trucks crossing the count

stations.

Data on weekly heavy truck traffic volumes (for May 9–15, 2010) were extracted from

FDOT’s TTMS traffic counting data. Specifically, data on the total weekly volume of heavy

trucks (i.e., class 8–13 trucks) were extracted. Figure 5.2 shows the TTMS locations in Florida

and the range of heavy truck traffic volumes (average daily traffic volumes) at those locations.

Highest tuck traffic volumes (i.e., around 5,000 heavy trucks per day) can be observed in the

northern part of I-75 near and above Ocala and on I-95 near Jacksonville. The section of I-75

53

between Ocala and Tampa, I-4 between Tampa and Orlando, and the section of I-95 in southeast

Florida have heavy truck volumes in the range of 3,000 to 4,000 trucks per day.

Figure 5.2 Observed Heavy Truck Traffic Flows at Different Telemetered Traffic Counting

Sites (TTMS) in Florida

Out of more than 250 TTMS traffic counting locations in Florida, only 160 had traffic

count data for all seven days in the week. Therefore, only these locations were selected for

further analysis to compare the ATRI truck traffic volumes with observed TTMS truck traffic

volumes at these locations. Figure 5.3 shows the results for each individual TTMS station.

Specifically, the blue bars in the figure represent the observed heavy truck traffic volumes at

those locations, and the red bars represent the ATRI truck traffic volumes through those

locations. Clearly, at no single location did the ATRI data provide 100 percent coverage of the

observed truck traffic flows. However, the data did provide some coverage of the heavy truck

traffic flows at all locations.

54

Figure 5.3 Heavy Truck (Classes 8–13) Counts from TTMS Data vs. Truck Counts

from ATRI Data during May 9–15, 2010

Table 5.1 shows these results aggregated by the facility type. The second column in the

table shows the number of TTMS traffic counting locations on different types of highway

facilities (along with the corresponding percentages). The third column shows the observed

TTMS volumes counted at these sites (along with the corresponding percentages), again,

separately for each highway facility type. Notice that a bulk of heavy truck traffic (65.6%) was

through locations on freeways and expressways that represent only 18.1 percent of the 160

TTMS sites considered in this analysis. The fourth column shows the truck traffic volumes

counted in ATRI data using the previously-discussed map-matching procedure. As can be

observed from the last row in this column, 163,467 ATRI truck crossings were counted at the160

TTMS locations. It is worth noting that the distribution of these ATRI truck traffic counts across

different facility types was very similar to the distribution of TTMS truck counts across facility

types. This can be observed by comparing the percentage numbers in the third and fourth

columns. This result suggests that the ATRI data provide a representative coverage of truck

flows through different facility types in the state. The last column expresses the ATRI truck

traffic counts as a percentage of observed heavy truck traffic counts at the TTMS locations. For

example 111,608 ATRI truck crossings were counted at TTMS locations on freeways and

expressways. These constitute 10.5 percent of more than 1 million observed heavy truck traffic

counts at these locations. These percentages provide an aggregate picture of the coverage

provided by ATRI data of the heavy truck traffic flows in Florida. Overall, as can be observed

from the last row in the last column of the table, it can be concluded that the ATRI data

provided 10.1 percent coverage of heavy truck flows observed in Florida. This result was

useful in many ways. First, this provided an idea of the extent of coverage of Florida’s heavy

truck traffic flows in the ATRI data. Second, the result could be used to weigh the seed matrix of

ATRI truck trip flows (by 10.1 times) to create a weighted seed matrix that can be used as an

input for the ODME process.

55

Table 5.1 Aggregate Coverage of Heavy Truck Traffic Volumes in Florida

by ATRI Data (for One Week from May 9–15, 2010)

Facility Type

No. of TTMS

Traffic

Counting

Stations

Observed TTMS

Truck Traffic

Volumes (Classes 8–

13) during May 9–15,

2010

Truck Traffic

Volumes in

ATRI Data during

May 9–15, 2010

% Coverage

Assuming ATRI

Data Comprises

Trucks of

Classes 8–13

Freeways & Expressways 29 (18.1%) 1,063,765 (65.6%) 111,608 (68.3%) 10.5%

Divided Arterials 64 (40.0%) 333,791 (20.6%) 30,472 (18.6%) 9.1%

Undivided Arterials 52 (32.5%) 101,066 (6.2%) 6,969 (4.3%) 6.9%

Collectors 8 (5.0%) 42,164 (2.6%) 5,127 (3.1%) 12.2%

Toll Facilities 7 (4.4%) 80,493 (5.0%) 9,291 (5.7%) 11.5%

Total 160 1,621,279 163,467 10.1%

5.4 Geographical Coverage of ATRI’s Data in Florida

To understand the geographical coverage of ATRI’s data in Florida, the number of trips

originating from (i.e., trip productions) and the number of trips destining to (i.e., trip attractions)

each traffic analysis zone (TAZ) of the FLSWM were plotted. Figure 5.4 shows the TAZ-level

trip productions and attractions, and Figure 5.5 shows the county-level trip productions and

attractions. Note that these trip productions and attractions were derived using the truck trips

derived from four months of ATRI’s truck GPS data. Since the trips were derived from four

months of data (i.e., 122 days), the total trip productions and attractions derived from the data

were first divided by 122 to get average daily trip productions and attractions. However, since

the data were found to represent 10 percent of observed heavy truck traffic volumes in the state,

the average daily trip productions and attractions were weighted by 10. Such weighted average

daily trip productions and attractions are shown in Figures 5.4 and 5.5.

In Figure 5.4, the TAZs shaded in yellow color have zero trip productions (left side of

figure) or zero trip attractions (right side) in ATRI’s four-month GPS data. It can be observed

that the Everglades region in the south and some TAZs in the northwest part of Florida have

TAZs with no trips extracted from the data. It was reasonable that zero to limited heavy truck

trips are produced from or attracted into the Everglades region. However, it was not clear if zero

trip generation in some TAZs of northwestern Florida was a result of low penetration of data in

those regions or if those TAZs have no freight truck trip flows. To investigate this further, the

observed heavy truck traffic flows in the TTMS data (Figure 5.2) can be examined. Except on I-

10, the northwestern region of the state did not have high truck traffic volumes. This suggests

that zero trip generations in the ATRI data for several TAZs in the northwestern region was a

reasonable representation of the truck flows in that region. Some TAZs in Duval (Jacksonville

area), Putnam, Polk, and Desoto counties had higher trip generation according to the ATRI data.

When examined closely, all these TAZs had major freight activity centers, such as distribution

centers. However, major urban areas such as Miami, Tampa, and Orlando did not show TAZs

with high trip generation. This was likely because the TAZs in these regions were smaller in size.

Since the trip generations were not normalized by the area of the TAZ, it was difficult to make

further inferences on the reasonableness of the TAZ-level trip generations.

56

Figure 5.4 FLSWM TAZ-Level Trip Productions and Attractions in the ATRI Data (4 Months of Data Factored to One Day)

57

Figure 5.5 County-Level Trip Productions and Attractions in the ATRI Data (4 Months of Data Factored to One Day)

58

Figure 5.5 shows the trip generations aggregated to a county-level. According to the

ATRI data, Duval, Polk, Orange, Miami-Dade, and Hillsborough counties, in that order, had the

highest truck trip generation. It was expected that counties with major metropolitan areas have

the highest heavy truck trip generation. Further, Polk County was expected to have a high truck

trip generation due to the presence of several freight distribution centers in the county. However,

it was interesting that the truck trip generation in Polk County was higher than that in

Hillsborough (Tampa), Orange (Orlando) and Miami-Dade counties. Further, the truck trip

generation in southeast Florida (Miami-Dade, Broward and Palm Beach counties) appeared to be

smaller than that in Polk County. These trends were not expected and were likely to be a

manifestation of spatial biases in the data. To address such spatial biases, Chapter 6 combines the

truck trip flows derived from the ATRI data with observed heavy truck traffic volumes at

different locations in the state.

Figure 5.6 presents the percentage of heavy truck traffic covered by ATRI data at

different locations. This information is similar to what was presented in Figure 5.3. However, for

clarity and ease in interpretation, the percentage covered is presented only for those locations

with annual average daily heavy truck traffic greater than 1,000 trucks per day. It can be

observed that at most locations, at least five percent of the heavy truck traffic was captured in the

ATRI data. Also, it can be observed that the coverage in the southern part of Florida (within

Miami) and the southern stretch of I-75 was slightly lower compared to the coverage in the

northern and central Florida regions.

Figure 5.6 Percentage of Observed Heavy Truck (Classes 8-13) Volumes Represented by

ATRI Data at Telemetric Traffic Monitoring Sites in Florida during May 9–15, 2010

59

CHAPTER 6 : ESTIMATION OF STATEWIDE ORIGIN-DESTINATION

TRUCK FLOWS

6.1 Introduction

The freight truck trips derived from ATRI’s truck GPS data can be used to derive statewide

origin-destination tables (or OD flow tables or OD matrices) of freight truck flows between

various traffic analysis zones. However, it was important to note that whereas the trips derived

from the ATRI data were substantial, they represented only a sample and not a census of all

freight truck flows within, to, and from the state. In addition, while it was known that the ATRI

data comprised predominantly tractor-trailer trucks that fall under classes 8–13 of FHWA vehicle

type classification (i.e., heavy trucks), it was not certain if the data represented a random sample

of heavy truck flows in the state. Therefore, additional information and procedures must be

employed to factor the sample of trips derived from the ATRI data to represent the population of

heavy truck flows within, to, and from the state. The weighting process was required not only for

inflating the sample of ATRI truck flows to the population truck flows but also for ensuring that

the spatial distribution of the resulting truck flows was representative of the actual truck flows in

the state. One approach to do this was Origin-Destination Matrix Estimation (ODME), which

involved combining the sample OD truck flows derived from the ATRI data with other sources

of information on truck flows observed at various links of the highway network to estimate a full

OD flow matrix representing the population of truck flows in the state. This chapter describes the

ODME approach used in this project, along with the results and findings. Section 6.2 describes

the mathematical procedure of the ODME approach used in this study, Section 6.3 describes the

inputs and assumptions in the ODME procedure, Section 6.4 presents the results, and Section 6.5

provides some suggestions to improve the ODME procedure and results.

6.2 Origin-Destination Matrix Estimation (ODME)

ODME is a class of mathematical procedures used to update an existing matrix of OD trip flows

(i.e., number of trips between each origin-destination pair in a study area) using information on

traffic flows at various locations in the transportation network. In the current project, the sample

OD truck flows (also called the sample OD matrix or the seed matrix) extracted from ATRI’s

truck GPS data can be updated using external information on truck traffic flows (or traffic counts

or traffic volumes) observed on various links in the highway network within and outside Florida.

Very broadly, the ODME procedures factor the ATRI data-derived truck trip flows in such a way

that the trips in the resulting estimated OD flow matrix, when assigned to the highway network,

closely match the observed heavy truck counts at various locations on the network.

The mathematical procedure used for ODME in this project was based on the ODME

procedure embedded in Cube Analyst Drive software from Citilabs. The procedure essentially is

an optimization problem that tries to minimize a function of the difference between observed

traffic counts and estimated traffic counts (from the estimated OD matrix) and the difference

between the seed matrix and the estimated OD matrix, as below:

0

a rg m in

su b jec t to 0 an d X

X

lo w er u p p er

J X F A X b G X X

X X X

(1)

60

In this general optimization problem, X is the OD matrix to be estimated, 0

X is the initial

(seed) OD matrix, G is a function measuring the distance between the estimated OD matrix and

the initial matrix, b is a vector of observed counts at different locations in the study area, A is

the route choice probability matrix obtained from the assignment of the OD flows in X on to the

network ( AX represents the estimated traffic counts at the same locations with available

observed counts), and F is a function measuring the difference between estimated and observed

traffic counts at different locations in the study area. As can be observed, the ODME procedure

attempts to arrive at an OD flow matrix X in such a way that the resulting traffic volumes at

different locations ( AX ) match closely with the observed traffic flows ( b ). At the same time,

the procedure avoids overfitting to the observed traffic flows by including the term 0G X X

so that the estimated matrix is not too far from the seed matrix. Xlo w e r

and u p p e r

X are boundaries

(lower and upper bounds) within which the estimated matrix should fall. The analyst has an

option to use these boundary constraints to set lower and upper bounds on the estimated matrix,

relative to the seed matrix.

Figure 6.1 shows a schematic of the ODME procedure used in the project. The primary

inputs to the procedure were the seed matrix for freight truck trips (derived from the ATRI data),

a highway transportation network for the study area and information on the travel times and

capacity of each link in the network (extracted from the FLWSM), and observed heavy truck

traffic volumes (or counts) at different locations added to corresponding links in the network. In

addition to these, OD flow matrices corresponding to travel other than freight truck flows—an

OD matrix for non-freight truck trips and an OD matrix for passenger travel (both extracted from

the FLSWM)—were required as inputs to generate realistic travel conditions in the network.

In the first step of the ODME procedure, the seed matrix of truck trips derived from the

ATRI data (assumed to represent a sample of freight truck trips) and other OD matrices

representing passenger travel and non-freight truck travel were loaded on to the highway

network using user-equilibrium-based traffic assignment procedures. The freight truck traffic

volumes estimated from the traffic assignment procedure were then used in conjunction with the

heavy truck traffic volumes observed at different locations in the network (along with the seed

and estimated OD matrices for freight trucks, which were same in the first iteration) to compute

the ODME objective function to be minimized. The seed matrix for freight truck trips was then

updated toward minimizing the objective function while considering the lower and upper bounds

on the matrix. This updated matrix was then used in conjunction with other OD matrices for

passenger and non-freight travel (which were not updated in the procedure) as the seed matrix

for the next iteration of the ODME procedure, which begins with highway traffic assignment and

follows with the computation of the ODME objective function. This process was repeated until

the ODME objective function reached its minimum, when the estimated freight truck traffic

volumes were close enough to observed volumes and the estimated OD matrix was not too far

from the initial seed matrix derived from the ATRI data.

61

Figure 6.1 Schematic of the ODME Procedure Used in This Project

The estimated OD matrix of freight truck trips can be evaluated using different evaluation

metrics and procedures. One approach to evaluate the estimated OD matrix is based on

comparison of estimated heavy truck traffic volumes and observed heavy truck traffic volumes at

different locations within and outside Florida. Specifically, one can evaluate a root mean square

error (RMSE) measure as below:

2

1

N

i i

i

a v g

V C

NR M S E

C

(2)

where, i

V is the estimated heavy truck traffic volume corresponding to location i, Ci is the

observed heavy truck traffic volume corresponding to location i, Ca v g

is the average heavy truck

traffic count value of the entire set of observations, and N is the total number of truck counting

locations in the set. A single RMSE value can be computed for the entire set of traffic counting

locations and also separately for locations in Florida and for locations elsewhere. Similarly, the

RMSE values can be computed separately for different ranges of observed heavy truck counts.

62

In addition to comparing the observed and estimated traffic volumes, it was important to

assess the reasonableness of the estimated OD matrix in different ways. Aggregating the OD

matrix to a coarser spatial resolution and examining the spatial distribution of flows, examining

the total trips originating from (or trip productions) and total trips destined to (or trip attractions)

each aggregate spatial zone, and examining the trip length distribution of the estimation OD

matrix in comparison to the seed OD matrix were different ways of assessing the reasonableness

of the estimated OD matrix.

6.3 Inputs and Assumptions for the ODME Procedure

6.3.1 The Seed Matrix

The seed matrix is essentially the matrix of OD truck trip flows derived from ATRI’s truck GPS

data. Specifically, the truck trips derived from the GPS data were assigned to the TAZ system

used in the FLSWM to form the seed matrix. In FLSWM, Florida and the rest of the United

States (and Canada) are divided into 6,242 TAZs; 5,403 of these zones are in Florida and the

remaining zones are outside Florida. Therefore, the seed matrix is a matrix of size 6242 × 6242,

with each cell in it representing the number of trips extracted between the corresponding origin-

destination (OD) pair.

In this project, the seed matrix was derived from four months of truck GPS data—March,

April, May, and June 2010. As described in an earlier chapter, whereas the total number of trips

derived from four months of data was more than 2.7 million, for the purpose of OD matrix

estimation, only those trips deemed to be made by heavy trucks that haul freight (i.e., FHWA

classes 8–13 trucks, which are tractor-trailers) were considered here. This is because most freight

in the U.S. is carried by tractor-trailer trucks of five axles or more10

(i.e., class 9 or above), and

some on tractor-trailer units of fewer than five axles (i.e., class 8 trucks). From discussions with

ATRI, whereas most of the ATRI data comprise tractor-trailer trucks, it is known that a small

share of trucks in the data do not necessarily haul freight over long distances. However, the raw

data did not provide information on the classification of each truck. Therefore, some heuristics

were developed to filter out trucks of class 7 or below. The heuristics used are discussed next.

Since the raw GPS data for each truck were available for at least two weeks (up to one

month, in most cases), trucks that did not make at least one trip of 100-mile length in a two-week

period were removed from the data. In this step, 88,869 trips made by 7,018 unique truck IDs

were removed. The median length of such removed trips was 20 miles, suggesting the short-haul

nature of these trucks. Subsequently, trucks that made more than five trips per day were

removed, assuming that these trucks are not freight carrying, tractor-trailer combination trucks.

In this step, 275,224 trips made by 918 unique truck IDs were removed. The median length of

these trips was 16 miles. The remaining trucks in the data were considered to be tractor-trailer

combination trucks that tend to make long-haul, freight carrying trips of interest to the FLSWM.

Among the trips made by these tractor-trailer combination trucks, the following three scenarios

were considered:

10 http://www.fhwa.dot.gov/policy/vius97.pdf.

http://www.fhwa.dot.gov/policy/vius97.pdf

63

1) Only trips of greater than 10 miles made by tractor-trailer combination trucks,

2) Only trips of greater than 5 miles made by tractor-trailer combination trucks, and

3) All trips made by tractor-trailer combination trucks.

After the above discussed procedures, the OD matrix of truck trips derived from 4

months of data comprised 2.07 million trips. Since this was derived from 4 months of data (122

days), the OD matrix was scaled down to one day by dividing all the cells in the OD matrix by

122. The resulting OD matrix then represents one day of trips extracted from the ATRI data. We

call this the one-day seed matrix. However, as discussed in Chapter 4, the research team found

that at an aggregate level, the trucks flows in the ATRI data represent 10 percent of the heavy

truck flows in Florida. Therefore, the one-day seed matrix was inflated by multiplying all the

cells in the matrix by 10. We call this the weighted one-day seed matrix (or seed matrix). This

weighted one-day seed matrix was the input as a seed matrix for the ODME process.

6.3.1.1 Geographical Coverage of the Seed Matrix

The spatial structure of the seed matrix is now examined, focusing on its geographical coverage.

To do so, the seed matrix was aggregated to county level within Florida and state level outside

Florida. The aggregation enabled meaningful analysis of the spatial coverage of the seed matrix.

Within the state of Florida, there are 67×67 = 4,489 county-to-county OD pairs. Of these,

the seed matrix derived from the ATRI data contained trips for 3,564 OD pairs (i.e., 79.4%

coverage). The remaining 925 (20.6%) of the county-to-county OD pairs in Florida did not have

trips in the seed matrix. From this, it can be concluded that the heavy truck trips derived from the

ATRI data covered close to 80 percent of the OD pairs in Florida. To examine the remaining 20

percent of OD pairs for which the seed matrix did not contain any trips, Table 6.1 separates those

OD pairs by county of origin (in the second column) and county of destination (in the third

column). For example, it can be observed from the row for the Baker County that 9 counties in

Florida did not have trips coming from the county and 7 counties did not have trips going into

the county. That is, the seed matrix contained trips coming from all other 58 (= 67 - 9) counties

to Baker and trips going from Baker to 60 (= 67 - 7) counties in Florida. A close examination of

this table suggested that counties associated with major urban regions (Miami-Dade,

Hillsborough, Orange, and Duval) and other counties with large freight activity (e.g., Polk) had a

small number of counties to or from which there were no trips in the seed matrix. Counties in

northwest Florida such as Franklin, Gulf, Calhoun, Holmes, Lafayette, Jefferson, and Hamilton

and some rural counties in the south such as Glades, Hardee, and Monroe had higher number of

zero trip flows coming into and going out of other counties in Florida.

64

Table 6.1 County-to-County OD Pairs in Florida with No Trips in the

Seed OD Flow Matrix Derived from ATRI’s Truck GPS Data

County No. of Counties to which There

are No Trips in Seed Matrix

No. of Counties from which

There are No Trips in

Seed Matrix

Alachua 0 1

Baker 9 7

Bay 10 4

Bradford 11 10

Brevard 8 9

Broward 5 4

Calhoun 33 37

Charlotte 19 17

Citrus 16 15

Clay 9 8

Collier 14 18

Columbia 5 5

De Soto 11 8

Dixie 18 14

Duval 0 2

Escambia 13 11

Flagler 18 20

Franklin 45 43

Gadsden 8 8

Gilchrist 19 20

Glades 30 32

Gulf 36 39

Hamilton 24 22

Hardee 27 26

Hendry 13 20

Hernando 9 8

Highlands 15 15

Hillsborough 1 2

Holmes 30 33

Indian River 19 24

Jackson 6 10

Jefferson 33 35

Lafayette 35 34

Lake 3 3

Lee 7 8

Leon 4 7

Levy 19 13

Liberty 21 7

Madison 8 5

Manatee 5 7

Marion 4 3

Martin 16 16

65

Table 6.1 (cont.)

County No. of Counties to which There

are No Trips in Seed Matrix

No. of Counties from which


Seed Matrix

Miami-Dade 3 3

Monroe 31 38

Nassau 6 2

Okaloosa 19 10

Okeechobee 20 24

Orange 2 1

Osceola 2 5

Palm Beach 6 7

Pasco 5 9

Pinellas 5 7

Polk 1 0

Putnam 4 4

Santa Rosa 23 20

Sarasota 9 14

Seminole 8 16

St. Johns 11 13

St. Lucie 7 8

Sumter 4 8

Suwannee 9 8

Taylor 12 12

Union 20 19

Volusia 6 4

Wakulla 22 23

Walton 24 25

Washington 30 15

Total 925 925

For OD pairs with at least one end in Florida, Table 6.2 presents the number of OD pairs

with no trips for each origin and destination county in Florida. Specifically, the second column

shows the number of states outside Florida to which no single trip was extracted, and the third

column shows the number of states from which no single trip was extracted. For example, it can

be observed from the row for the Alachua County that 11 states outside Florida did not have trips

coming from the county and 10 states did not have trips going into the county. Similar to the

county-to-county flows, counties in the northwest such as Franklin, Gulf, Calhoun, Hamilton,

Walton, Holmes, Lafayette, and Jefferson and rural counties in the south such as Glades and

Monroe had no trips coming into or going out of other states.

66

Table 6.2 Florida County to Non-Florida State OD Pairs with No Trips in the

Seed OD Flow Matrix Derived from ATRI’s Truck GPS Data

County

No. of Counties to which

There are No Trips in Seed

Matrix

No. of Counties from which There

are No Trips in

Seed Matrix

Alachua 11 10

Baker 20 11

Bay 9 11

Bradford 22 23

Brevard 15 11

Broward 8 5

Calhoun 41 37

Charlotte 26 21

Citrus 26 24

Clay 20 13

Collier 21 20

Columbia 22 19

De Soto 19 10

Dixie 26 34

Duval 4 2

Escambia 11 12

Flagler 25 23

Franklin 42 41

Gadsden 16 20

Gilchrist 23 28

Glades 33 33

Gulf 38 39

Hamilton 35 36

Hardee 26 27

Hendry 13 25

Hernando 17 18

Highlands 18 23

Hillsborough 4 4

Holmes 36 32

Indian River 18 24

Jackson 21 23

Jefferson 34 34

Lafayette 39 38

Lake 4 5

Lee 23 20

Leon 25 20

Levy 23 26

Liberty 25 32

Madison 23 19

Manatee 7 14

Marion 13 9

Martin 21 28

67

Table 6.2 (cont.)

County No. of Counties to which


Seed Matrix

No. of Counties from which There

are No Trips in

Seed Matrix

Miami Dade 5 2

Monroe 36 37

Nassau 16 18

Okaloosa 18 18

Okeechobee 24 26

Orange 4 3

Osceola 12 10

Palm Beach 10 9

Pasco 17 14

Pinellas 6 10

Polk 3 2

Putnam 8 12

Santa Rosa 22 20

Sarasota 22 17

Seminole 12 13

St. Johns 18 17

St. Lucie 18 15

Sumter 16 17

Suwannee 21 26

Taylor 24 29

Union 28 34

Volusia 8 12

Wakulla 25 27

Walton 32 32

Washington 26 29

Total 1,334 1,353

Overall, it can be concluded that the ATRI data provided a sound geographic coverage of

trip flows within, to, and from Florida. Whereas several counties in northwest Florida and a few

rural counties in south Florida (e.g., Glades and Monroe) showed no trips to and from several

other counties and states, it is likely because these counties may not actually have truck flows

to/from a large number of locations. Considering that the seed matrix was derived from four

months of raw GPS data (which is a large amount of data), if some OD pairs at a county-level

resolution did not have any trip exchanges, it is reasonable to expect that those OD pairs may not

have truck flows, in reality. On the other hand, for OD pairs with both ends outside the state of

Florida, 350 of the 2,500 (=50×50) state-to-state OD pairs did not have any trips in the seed

matrix. Since the data were Florida-centric, it was likely that the seed matrix was not necessarily

a good representation of OD flows outside Florida.

6.3.1.2 Zero Cells in the Seed Matrix

When the seed OD matrix was examined at its actual spatial resolution (i.e., the FLSWM TAZ

level), only 0.41 million of the 39 million TAZ-to-TAZ OD pairs had trips. That is, the 2 million

heavy truck trips extracted from ATRI’s truck GPS data could fill only 0.41 million OD pairs.

68

The remaining 38.5 million OD pairs had no trips. This is relevant because most ODME methods

used in practice operate only with OD pairs that have non-zero trips in the seed matrix.

Consequently, the final OD matrix output from ODME methods will have zero trips for OD pairs

that began with zero in the seed matrix. To address this issue, a common practice is to introduce

a small positive number (e.g., 0.01) for zero-cells (i.e., OD pairs with zero trips) in the seed

matrix that the analyst believes should have trip flows. The question, then, becomes which OD

pairs with zero trips can be expected to have trip flows in reality. The earlier discussion on the

spatial coverage of the seed matrix, albeit at an aggregate spatial resolution of counties and

states, sheds light on this issue. Recall from the earlier discussion that the OD pairs with at least

one end in Florida had sufficient coverage at the county level in Florida and at the state level

outside Florida. While there may be gaps at the disaggregate TAZ level, it was considered

unnecessary to alter zero cells for such OD pairs. For OD pairs outside Florida, it may be

reasonable to explore altering the zero-cells to include a small number (0.01) and examine if the

ODME procedure provides better results.

The following scenarios were considered for altering the zero cells in the seed matrix:

1) None of the zero cells were altered (assuming the zero cells in the seed matrix are

truly representative of zero truck flows),

2) Only the zero-cells for OD pairs outside Florida were altered to 0.01 to allow for the

possibility of truck flows between those OD pairs. This scenario assumes that zero

cells for OD pairs within, to, and from Florida are truly representative of zero truck

flows, and

3) All zero-cells were altered to 0.01. This scenario—that all OD pairs will have truck

flows—is very unlikely in reality. Nevertheless, it was considered to be sure.

6.3.2 Truck Traffic Counts

Observed volumes of truck traffic at different locations on the network was an important input

into the ODME process. For the current study, data on heavy truck traffic counts were gathered

for several locations within and outside Florida. Since the OD matrix to be estimated included

truck traffic flows going into (out of) Florida from (to) other states, it was considered important

to include truck traffic counts outside Florida as well.

6.3.2.1 Truck Traffic Counts in Florida

Data on truck traffic counts in Florida were obtained from FDOT’s Telemetered Traffic

Monitoring Sites (TTMS) traffic counting program. FDOT collects daily data on traffic volumes

(by direction), speed, vehicle type, and weight from more than 250 TTMS locations on Florida’s

highway network. From such TTMS data for the year 2010, the daily traffic volume information

for different vehicle classes was extracted for the months of March, April, May, and June 2010

(the same months for which the seed matrix is available). The vehicle classifications range from

1 to 15, with classes 8–13 representing heavy trucks (i.e., tractor-trailer combinations) and class

15 representing “Unknown”. For each TTMS location, the average daily traffic (ADT) was

computed for heavy trucks along with the number of days for which the traffic count data were

available. Subsequently, the data were examined for any anomalies as discussed below.

69

First, 241 locations whose coordinates fell on the FLSWM highway network locations

were selected. The other TTMS locations that were on highway links not in the FLSWM network

were removed from consideration. For 237 of these locations, the TTMS traffic count data were

available for both directions (141 sites with traffic counts in north–south directions and 96 sites

in east–west directions); the remaining 4 sites had counts for one direction. Thus, 478 TTMS

traffic counts were distinguished by location and direction. Of these, only 460 locations with

TTMS data for more than 30 days of the 4 months were considered. Subsequently, sites with the

following types of anomalies were removed: those with abnormally high percentage of traffic

counts, those with abnormally high difference in directional counts, and those with a high

percentage of unclassified trucks (i.e., class 15). After all these screening procedures, TTMS

heavy truck counts (i.e., ADT for heavy trucks) for 413 different locations were retained for use

in the ODME process. Figure 6.2 shows the spatial distribution of those locations in Florida

(percentages in parentheses show the distribution of heavy truck ADT at these 413 locations). It

is worth noting that 22.5 percent of these locations were on freeways and expressways, 36

percent were on divided arterials, 30 percent were on undivided arterials, 6.5 percent were on toll

facilities, and the remaining were on collector roads, ramps, one-way facilities, and centroid

connectors. Of the 413 heavy truck counts at different locations in Florida, data from 365

locations were used in the ODME process, and data from the remaining 48 locations were kept

aside for validation purposes.

Figure 6.2 Spatial Distribution of Telemetered Traffic Monitoring Sites (TTMS) in Florida

Used for ODME

70

6.3.2.2 Truck Traffic Counts outside Florida

FHWA’s Vehicle Travel Information System (VTRIS) database was used to obtain truck traffic

counts on highway network locations in all states other than Florida and Georgia. For Georgia,

truck traffic counts from Georgia Automated Traffic Recorder (ATR) locations were obtained

from Georgia Department of Transportation (GDOT). Whereas the VTRIS database provided

traffic count data for a large number of locations outside Florida and the Georgia ATR data

provided so similar data in Georgia, only 635 of these locations fell on the FLSWM highway

network links outside Florida. This is because the FLSWM network outside Florida is not very

detailed. Figure 6.3 shows the locations of all 635 counting sites along with the 413 counting

sites in Florida. As can be observed, Florida and Georgia had good coverage of traffic counting

stations, but other states in the southeast, such as Alabama, Mississippi, Louisiana, and South

Carolina, had very few traffic counting locations. Tennessee, Kentucky, and North Carolina did

not have any traffic counting locations. This will likely have a bearing on ODME results. In the

future, the ODME results potentially can be improved by increasing the spatial coverage of the

traffic counting stations in the southeastern states. Finally, of the 635 different heavy truck

counts at different locations outside Florida, data from 598 locations were used in the ODME

process and data from the remaining 37 locations were kept aside for validation purposes.

Figure 6.3 Spatial Distribution of Traffic Counting Stations in U.S. Used for ODME

71

6.3.3 Network

The highway network from the FLSWM was used as an input for traffic assignment purposes in

ODME. The network is detailed within Florida and just outside the border of Florida and less

detailed as it extends into other states farther from Florida. As can be observed from figure 6.3,

the network extends into Canada and Mexico as well. The inputs associated with the network

included the following:

a) Free flow speed, travel time (including any delays due to toll plazas), and capacity of

each link in the network, and

b) Observed average daily heavy truck counts by direction on over 200 links obtained

from TTMS data in Florida and VTRIS data in other states (figures 6.2 and 6.3 show

these traffic counting locations of those traffic counting locations).

6.4 Results from the ODME Procedure

6.4.1 Evaluation of Different Assumptions

The ODME procedure in CUBE’s Analyst Drive was run several times to evaluate different

assumptions on the OD matrix. These assumptions include assumptions (or constraints) of lower

and upper bounds on the seed matrix, assumptions on zero cells in the seed matrix, and

assumptions on the minimum trip length to be considered in the seed matrix.

Assumptions on the bounds for trips between different OD pairs in the seed matrix

include:

1) Lower bound equal to the seed matrix; i.e., none of the estimated number of trips

between any OD pairs should be less than those in the seed matrix,

2) No lower bound on the seed matrix; i.e., the estimated matrix can have zero trips

between OD pairs even if there were trip flows observed in the seed matrix,

3) Lower bound equal to 0.7 of the seed matrix; i.e., none of the estimated trips between

any OD pair should be less than 0.7 times those in the seed matrix,

4) Upper bound of 50 times the seed matrix; i.e., none of the estimated trips between any

OD pair should be more than 50 times those in the seed matrix,

5) Upper bound of 100 times the seed matrix; i.e., none of the estimated trips between

any OD pair should be more than 100 times those in the seed matrix, and

6) No upper bound on the seed matrix.

It is worth noting here that Cube’s Analyst Drive software does not allow the bounds to

be different across different cells in the matrix. The bounds have to be uniform across all cells.

Assumptions on zero cells in the seed matrix include the following:

1) Assume OD pairs with zero cells in the seed matrix do not have truck flows in reality

(i.e., all zero-cells were retained as zeroes),

2) Alter only the zero cells for OD pairs outside Florida to 0.01 to allow the possibility

of truck flows between those OD pairs, and

72

3) Alter all zero cells to 0.01, assuming that each zero-cell in the seed matrix is likely to

have truck trips in reality.

Assumptions on minimum trip length in the seed matrix include:

1) Minimum trip length of 10 miles,

2) Minimum trip length of 5 miles, and

3) Minimum trip length of 1 mile.

Among the assumptions on lower/upper bounds on the OD matrix, the extent of upper

bound did not influence the results (i.e., the estimated OD matrix) as long as the bound was large

enough. Therefore, the upper bound on the OD matrix was removed. The extent of lower bounds

had a significant influence on the estimated OD matrix. When lower bounds were removed on all

cells in the OD matrix, the estimated truck traffic volumes matched better than the scenarios that

imposed lower bounds on the OD matrix. This can be observed from Table 6.3. Specifically, the

RMSE values between estimated and observed truck traffic volumes were smallest when no

lower bounds were imposed on the OD matrix. However, in this scenario, the trip length

distribution of the estimated OD matrix was changing considerably toward a greater share of

shorter trips than those in the seed matrix. There were two possible reasons for such a change in

the trip length distribution from the seed matrix to the estimated OD matrix: one was that the

seed matrix was biased toward long-distance trips and that combining the seed matrix with the

observed traffic counts helped reduce the bias by estimating more short-length trips, and the

other was that the estimated OD matrix from the ODME procedure was over-fitting to the

observed traffic counts without necessarily correcting for biases in the seed matrix. To

investigate this further, any possible anomalies in the estimated OD matrix were closely

examined.

Table 6.3 RMSE Values between Estimated and Observed Heavy Truck Traffic Volumes

at Different Locations in Florida for Different Assumptions in ODME Procedure No Lower Bounds

Assumed on

OD Matrix

Lower Bound Equal

to No. of Trips in

Seed Matrix

Lower Bound Equal

to 0.7 Times No. of Trips

in Seed Matrix

Observed

Daily Heavy

Truck Counts

RMSE for

Input

Stations

RMSE for

Validation

Stations

RMSE for

Input

Stations

RMSE for

Validation

Stations

RMSE for

Input

Stations

RMSE for

Validation

Stations

20–100 8% 92% 83% 104% 70% 105%

100–500 6% 91% 77% 93% 47% 93%

500–1000 2% 38% 46% 47% 33% 49%

1000–7000 2% 25% 37% 40% 11% 24%

All 4% 37% 59% 60% 20% 38%

When closely examined, the estimated OD matrix (when the lower bounds were

removed) had zero trips between many OD pairs that originally had some observed trips in the

seed matrix from the ATRI data. While this was not necessarily a problem in itself, the estimated

OD matrix had zero trips between Florida and some southeastern states that had no observed

traffic counts from the VTRIS data (recall that no observed truck traffic counts from Tennessee

73

and North Carolina could be used). For example, no OD pair between North Carolina and

Florida and between Tennessee and Florida had any trips in the estimated OD matrix, although at

least 100 trips were observed between those states and Florida in the ATRI data. This suggested

that the estimated OD matrix was an artifact of over-fitting to the observed traffic volumes rather

than a realistic representation of OD flows within, to, and from Florida. A closer examination of

the RMSE values for this scenario also suggested that the ODME procedure in this scenario was

over-fitting to the observed traffic counts. Specifically, the RMSE value between the estimated

and observed heavy truck volumes for input stations (i.e., the TTMS locations from which the

data were used for ODME procedure) was only 4 percent. Such an excellent fit to the observed

data did not translate to the validation data; i.e., the RMSE between estimated and observed

heavy truck volumes was 37 percent for TTMS locations from which the truck traffic count data

was kept aside for validation. Therefore, the research team believes that the estimated OD matrix

in this scenario was an artifact of over-fitting to the observed truck traffic volumes. In future

work, this issue can be resolved by obtaining better observed truck traffic count information

from all southeastern states, especially those that do not have any or very few traffic counts in

the inputs used in the project.

When the lower bound was set to be equal to the seed matrix, the estimated OD matrix

was very close in its trip length distribution to the seed matrix. However, the heavy truck traffic

volumes implied by the estimated OD matrix (obtained from traffic assignment) were not close

enough to the observed heavy truck traffic counts. The root mean squared value between the

estimated traffic volumes and the observed traffic volumes was close to 60 percent.

As a middle ground between the above two scenarios, a scenario was explored in which

the lower bounds were set to be 0.7 times the seed matrix (i.e., none of the estimated trips

between any OD pairs should be less than 0.7 times those in the seed matrix). Note that the seed

matrix used as input for the ODME procedure was a 10-fold inflated version of the one-day seed

matrix extracted from the ATRI data. This was done to recognize that the ATRI data represented

about 10 percent of the observed heavy truck flows in the state (at an aggregate level). However,

it is not necessary that the data represents 10 percent of heavy truck flows at every location. In

some locations, the data might represent more or less than 10 percent of the observed heavy

truck flows. Therefore, setting a lower bound of 0.7 allows for the possibility that the actual

heavy truck trip flows might be less than the 10-fold inflated number of heavy truck trips in the

ATRI data. This scenario provided reasonable results, with RMSE value of 20 percent for input

stations and 38 percent validation stations while also allowing trips from (and to) all states to

(and from) Florida.

Among the assumptions on zero cells, keeping the zero cells as is provided better results

both in terms of validation measures against observed heavy truck counts as well as

reasonableness of the spatial distribution of truck flows. For instance, altering all zero cells to

0.01 provided high RMSE values results unless the lower bounds were removed on all cells.

However, removing the lower bounds on all cells, as discussed earlier, was leading to over-

fitting of the estimated heavy truck traffic volumes to observed truck traffic volumes. Since there

was no easy mechanism in Cube’s Analyst Drive software to incorporate different bounds for

different OD pairs, lower bounds could not be imposed on only non-zero cells in the OD matrix

and allow the altered zero cells to become zero. Besides, since the seed matrix was derived using

74

a large database from four months of ATRI data, zero cells can be reasonably assumed to

represent no truck flows between the corresponding OD pairs. Recall from the discussion in

Section 6.3.1.1 that when the seed matrix was aggregated to the county level, few OD pairs in

the state had zero trips.

Assumptions on minimum trip length in the seed matrix did not significantly alter the

estimated OD matrix except that assumptions with smaller trip length cutoffs led to a higher

share of intra-county trips. Since the purpose of this effort was toward statewide freight truck

flow modeling, we retained the assumption that valid pickup/delivery trips of heavy trucks

should be of at least 10-mile length.

6.4.2 ODME Results for One Set of Assumptions

This section presents and discusses results from the following set of assumptions in the ODME

procedure: (1) no upper bounds, but a lower bound of 0.7 times the seed matrix on the estimated

OD matrix; (2) trips of at least 10-mile length; and (3) zero cells in the seed matrix assumed to

truly represent zero truck flows. The results based on these assumptions were considered to be

the final results in the project. However, there is scope for improving the results, which is

discussed toward the end of this chapter.

Table 6.4 presents a summary of the truck trips in the seed matrix and those in the

estimated OD matrix. As mentioned previously, the seed matrix had trips between nearly 0.41

million OD pairs. Of these, close to 0.3 million OD pairs had at least one end in Florida, and 0.18

million OD pairs had both ends in Florida. As can be observed from the table, the same OD pairs

had trips in the estimated OD matrix. The seed matrix contained a total of 69,025 trips that

started and/or ended in Florida, and the estimated OD matrix contained a total of 104,587 trips.

Close to 70 percent (73,202) of the estimated trips with at least one end in Florida were within

Florida. The daily mileage of estimated trips with at least one end in Florida was more than 27

million miles. A total of 26.6 percent of these miles (more than 7 million miles) was due to trips

within Florida.

Table 6.4 Summary of Truck Trips in Seed and Estimated OD Matrices Seed OD Matrix Estimated OD Matrix

Trips between All OD pairs within and outside Florida

No. of OD pairs with trips 410,559 410,559

No. of trips per day 169,859 343,071

Miles traveled per day 54,642,638 206,760,073

Trips with at Least One End in Florida




Trips with Both Ends in Florida




75

Figure 6.4 shows a comparison of estimated truck traffic volumes (using the estimated

OD matrix) and observed truck traffic volumes in the TTMS data. The blue dots in the figure are

for TTMS stations from which the observed traffic volume data was used on the ODME process,

and the red dots are for TTMS stations from which the observed traffic volume data were kept

aside for validation. The solid straight line is the 45-degree line. All dots that fall on this line

indicate a perfect fit between the estimated truck volume and the observed truck volume. The

dots that fall between the two dotted lines correspond to those locations at which the estimated

truck traffic volumes are within 25 percent deviation from the observed truck traffic volumes. A

table embedded within the figure shows the aggregate RMSE values for different ranges of

observed truck traffic volumes. It can be observed that the estimated truck traffic volumes were

matching reasonably well with the observed volumes, especially at locations with truck volumes

higher than 1,000 trucks per day.

Figure 6.5 shows the trip length distributions of the trips in the seed and estimated OD

matrices (the trip lengths are based on TAZ-to-TAZ distances in the FLSWM). The top graph in

the figure shows the distribution for trips with at least one end in Florida (this include trips

between other states and Florida), and the bottom graph shows the distribution for trips with both

ends in Florida. It can be observed that the distribution of the trips in the estimated OD matrix

was closely following those from the seed matrix derived from the ATRI data, albeit the

estimated OD matrix had a slightly greater proportion of shorter length trips than the seed matrix.

Notice from the top graph that the estimated trips showed a spike in the trips of length greater

than 2,000 miles when compared to those in the seed matrix. These likely were trips between

Florida and states from the northwestern U.S., including California.

Figures 6.6 and 6.7 show the county-level trip productions and attractions, respectively,

for both the seed and estimated OD matrices. As discussed in Chapter 4, the seed matrix showed

lower than expected trip generation in the south Florida region (especially in and around Miami)

and the southern stretch of I-75 beginning from the Tampa region (when compared to those in

the Polk County). These trends were observed in discussions related to the coverage of heavy

truck flows in Florida by the ATRI data. The estimated OD matrix, due to its use of additional

information on the observed heavy truck traffic flows, addressed this issue to a certain extent.

This can be observed in Figures 6.6 and 6.7, where counties in southeast Florida and

Hillsborough County had higher trip generation in the estimated OD matrix than in the seed OD

matrix. Also note that the trip generation in the Duval County has increased as well. This was

perhaps due to a high volume of heavy truck traffic in the Jacksonville region.

76

0

1000

2000

3000

4000

5000

6000

7000

0 1000 2000 3000 4000 5000 6000 7000

Esti

mat

ed

he

avy

tru

ck c

ou

nts

pe

r d

ay

Observed heavy truck Counts per day

All Input Stations

Validation Stations

Linear (45 Degree Line)

Linear (25% Error Line)

Figure 6.4 Observed vs. Estimated Heavy Truck Counts per Day at Different Locations in Florida

Observed truck

counts per day

RMSE for input stations

(No. of input Stations)

RMSE for validation stations

( No. of validation Stations)

20-100 70% (64 locations) 105% (11 locations)




Total 20% (365 locations) 38% (48 locations)

77

Figure 6.5 Trip Length Distributions of Trips in Estimated and Seed OD Matrices

0

5

10

15

20

25

30

35

<1

0

50

-10

0

15

0-2

00

25

0-3

00

35

0-4

00

45

0-5

00

55

0-6

00

65

0-7

00

75

0-8

00

85

0-9

00

95

0-1

00

0

10

50

-11

00

11

50

-12

00

12

50

-13

00

13

50

-14

00

14

50

-15

00

15

50

-16

00

16

50

-17

00

17

50

-18

00

18

50

-19

00

19

50

-20

00

Pe

rce

nta

ge o

f tr

ips

wit

h a

t le

ast

on

e e

nd

in F

lori

da

Trip Length/Distance between TAZs (Miles)

ATRI Data (Seed Trips)

Estimated Trips

0

5

10

15

20

25

30

35

<1

0

10

-50

50

-10

0

10

0-1

50

15

0-2

00

20

0-2

50

25

0-3

00

30

0-3

50

35

0-4

00

40

0-4

50

45

0-5

00

50

0-5

50

55

0-6

00

60

0-6

50

65

0-7

00

70

0-7

50

75

0-8

00

80

0-8

50

Pe

rce

nta

ge o

f tr

ips

wit

h b

oth

en

ds

in F

lori

da

Trip Length/Distance between TAZs (Miles)

ATRI Data (Seed Trips)

Estimated Trips

78

Figure 6.6 Comparison of Trip Productions by County between Seed OD Matrix and Estimated OD Matrix

79

Figure 6.7 Comparison of Trip Attractions by County between Seed OD Matrix and Estimated OD Matrix

80

Figure 6.8 Comparison of Truck Trip Productions and Attractions between Seed and Estimated OD Matrices

Broward

Duval

Hillsborough

Lake

Marion

Miami-Dade

Orange

Palm Beach

Polk

Putnam

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

0 2000 4000 6000 8000 10000

Nu

mb

er

of

Trip

s A

ttra

cte

d t

o C

ou

nti

es

in F

lori

da

in E

stim

ate

d O

D

Mat

rix

(Sce

nar

io 1

)

Number of Trips Attracted to Counties in Florida in Seed OD Matrix

Data Trend Line 45 Degree Line

Broward

Duval

Hillsborough

Miami-Dade

Orange Palm Beach

Polk

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

0 2000 4000 6000 8000 10000

Nu

mb

er

of

Trip

s P

rod

uce

d f

rom

Co

un

tie

s in

Flo

rid

a in

Est

imat

ed

OD

M

atri

x (S

cen

ario

1)

Number of Trips Produced from Counties in Florida in Seed OD Matrix

Data Trend Line 45 Degree Line

81

Tables 6.5 and 6.6 show the state-to-state trip flows in the seed and estimated OD

matrices for a selected set of states. Specifically, Table 6.5 shows the distribution of the trips

starting in Florida, and Table 6.6 shows the distribution of trips ending in Florida. The seed

matrix showed that around 75 percent of the trips staring (ending) in Florida stay within (are

from within) Florida, and the estimated matrix adjusted this distribution to contain about 82

percent of those trips within Florida. The next top destinations (origins) for trips starting (ending)

in Florida include Georgia, Alabama, and California. It is interesting that California was one of

the top destinations (origins) for trips starting in Florida. The reasons for this are not clear and

need further investigation. One possibility is that such heightened flows between California and

Florida may be artifacts of the ODME procedure given the observed heavy truck traffic volumes

in the two states.

Tables 6.7 and 6.8 show the county-to-county trip flows in the seed and estimated OD

matrices for counties with the highest truck trip productions and attractions in the data. In both

the tables, cells with greater than 10 percent value are shaded in red and those with 5–10 percent

value are shaded in brown. Table 6.7 shows the destinations of heavy truck flows from counties

with highest trip production in Florida, and Table 6.8 shows the origins of heavy truck flows to

counties with highest trip attraction. Several observations can be made from these tables. First, as

expected, a good proportion of trips to/from each county were from/to within the county.

Second, the seed matrix showed Polk County as one of the major origins/destinations for trips

from/to other counties. The estimated OD matrix made adjustments to this trend for Miami-

Dade, Palm Beach, and Broward counties; specifically, the estimated OD matrix showed greater

flows between these three counties. Third, the estimated OD matrix showed smaller proportion

of flows between Hillsborough and Miami-Dade counties than that in the seed OD matrix. While

one would expect greater amount of flows between these two counties, the observed heavy truck

traffic volumes on major highways between these two counties were not high enough to support

this notion.

Finally, Figure 6.9 shows a comparison of the truck trip flows between seed and

estimated OD matrices. It can be observed that only those OD pairs with less than 100 trips in

the seed matrix have been modified in the estimated seed matrix.

6.5 Scope for Improvements to ODME Results

Some improvements can be implemented in the ODME procedure to obtain a better OD matrix

of truck flows within, to, and from the state. Observed heavy truck volumes at different locations

within and outside Florida comprise an important input to the ODME process. The research team

made several efforts to obtain heavy truck counts for several states outside Florida, but for

several states in the Southeast, the team could not obtain the data for a large number of locations.

Therefore, the ODME results potentially can be improved by increasing the spatial coverage of

observed traffic volume data in the southeastern states (Alabama, Louisiana, South Carolina,

North Carolina, Mississippi, and Tennessee). This will help capture the heavy truck flows more

accurately, especially the flows between other states and Florida.

Cube’s Analyst Drive software does not allow the lower and upper bounds on the OD

matrix to be different across different cells in the matrix (i.e., the bounds have to be uniform

across all cells). Allowing different boundary conditions on different types of OD pairs (e.g.,

82

separate boundary conditions for OD pairs within Florida and for those outside Florida) may help

improve the ODME results. Further, closely examining the accuracy of observed truck traffic

counts and specifying confidence levels on this data at different locations may help address

issues related to the inaccuracy of the observed truck traffic count data.

Table 6.5 Heavy Truck Trip Flows from Florida to Other States

(Greater than 0.5 % in Estimated Matrix)

Destination

State AL CA FL GA IL MI NJ Other Total

Seed OD Matrix from ATRI Data 4.6% 0.1% 75.8% 10.3% 0.3% 0.2% 0.3% 8.4% 100.0%

Estimated OD Matrix 2.8% 1.2% 82.1% 9.3% 1.0% 0.9% 1.1% 1.6% 100.0%

Table 6.6 Heavy Truck Trip Flows from Other States to Florida

(Greater than 0.5 % in Estimated Matrix)

Origin

State

Seed OD Matrix

from ATRI Data

Estimated OD

Matrix

AL 3.9% 2.8%

CA 0.1% 1.4%

FL 76.4% 82.6%

GA 9.4% 8.4%

IL 0.3% 1.1%

MI 0.1% 0.6%

NJ 0.3% 0.8%

Other 9.5% 2.3%

Total 100% 100%

Table 6.7 Destinations of Heavy Truck Trip Flows from Selected Counties in Florida

Destination

Origin Broward Duval

Hillsbor

ough

Miami-

Dade Orange

Palm

Beach Polk … …

Broward (Seed Matrix) 13.6% 5.4% 1.9% 18.5% 3.8% 10.4% 8.5% … … 100%

(Estimated Matrix) 16.9% 2.0% 0.3% 28.0% 0.3% 29.0% 2.2% … … 100%

Duval (Seed Matrix) 2.2% 16.0% 2.5% 2.7% 4.5% 1.5% 3.7% … … 100%


Hillsborough (Seed Matrix) 0.8% 5.6% 14.4% 2.2% 6.4% 0.7% 15.5% … … 100%


Miami-Dade (Seed Matrix) 10.9% 6.2% 3.0% 15.4% 4.7% 7.5% 7.5% … … 100%


Orange (Seed Matrix) 1.8% 9.1% 5.2% 3.1% 10.4% 2.1% 15.1% … … 100%


Palm Beach (Seed Matrix) 10.6% 7.7% 1.9% 10.2% 6.2% 10.9% 7.8% … … 100%


Polk (Seed Matrix) 2.7% 4.3% 8.5% 3.1% 8.2% 2.0% 16.7% … … 100%


83

Table 6.8 Origins of Heavy Truck Trip Flows to Selected Counties in Florida

Destination

Origin Broward Duval Hillsborough Lake

Miami-

Dade Orange

Palm

Beach Polk

Broward (Seed Matrix) 13.6% 1.7% 1.1% 1.6% 12.6% 1.8% 12.5% 2.6%

(Estimated Matrix) 19.9% 1.1% 0.3% 0.3% 24.3% 0.3% 27.2% 1.4%

Duval (Seed Matrix) 6.7% 15.9% 4.4% 1.8% 5.7% 6.5% 5.5% 3.5%


Hillsborough (Seed

Matrix) 1.5% 3.2% 14.8% 4.7% 2.7% 5.5% 1.6% 8.6%


Miami-Dade (Seed

Matrix) 16.4% 3.0% 2.6% 4.2% 15.8% 3.4% 13.5% 3.5%


Orange (Seed Matrix) 3.8% 6.3% 6.4% 13.7% 4.5% 10.7% 5.6% 10.1%


Palm Beach (Seed

Matrix) 8.9% 2.1% 0.9% 1.1% 5.8% 2.5% 11.0% 2.0%


Polk (Seed Matrix) 8.8% 4.5% 16.0% 19.7% 6.9% 12.8% 7.7% 16.9%


…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

…

Total (Seed Matrix) 100% 100% 100% 100% 100% 100% 100% 100%

Total (Estimated Matrix) 100% 100% 100% 100% 100% 100% 100% 100%

Figure 6.9 Comparison of Truck Trip flows between Each OD Pair in

Seed and Estimated OD Matrices

84

CHAPTER 7 : SUMMARY AND CONCLUSIONS

7.1 Summary

Freight is gaining increasing importance in transportation planning and decision making at all

levels of the government, particularly MPOs, states, and at the federal level. An accelerated

growth in the volume of freight shipped on American highways has led to a significant increase

in truck traffic. This has put enormous pressure on national highways, which impacts traffic

operations, safety, highway infrastructure, port operations, and distribution center operations.

Traffic congestion, in turn, impedes the speed and reliability of freight movement on the

highway system and leads to direct economic costs for producers and consumers, passenger

traffic congestion, safety issues, and environmental impacts.

As freight movement continues to grow within and between urban areas, appropriate

planning and decision making processes are necessary to mitigate the above-mentioned impacts.

However, a main challenge in establishing these processes is the lack of adequate data on freight

movements such as detailed origin-destination (OD) data, truck travel times, freight tonnage

distribution by OD pairs, transported commodity by OD pairs, and details about truck trip stops

and paths. As traditional data sources on freight movement are either inadequate or no longer

available, new sources of data must be investigated.

A recently-available source of data on nationwide freight flows is based on a joint

venture by American Transportation Research Institute (ATRI) and Federal Highway


performance measures (FPM) on key corridors in the nation. These data are obtained from

trucking companies who use GPS-based technologies to remotely monitor their trucks. ATRI’s

truck GPS database contains GPS traces of a large number of trucks as they traveled through the

national highway system. This provides unprecedented amounts of data on freight truck

movements throughout the nation (and Florida). Such truck GPS data potentially can be used to

support planning, operation, and management processes associated with freight movements.

Further, the data can be put to better use when used in conjunction with other freight data

obtained from other sources.

The overarching goal of this project was to investigate the use of ATRI’s truck GPS data

for statewide freight performance measurement, statewide freight truck flow analysis, and other

freight planning and modeling applications in the state. The specific objectives of the project

were to:

1) Derive freight performance measures for Florida’s Strategic Intermodal System (SIS)

highways,

2) Develop algorithms to convert large streams of ATRI’s truck GPS data into a more

useable truck trip format,

3) Analyze truck trip characteristics in Florida using ATRI’s truck GPS data,

4) Assess ATRI’s truck GPS data in terms of its coverage of truck traffic flows in

Florida,

85

5) Develop OD tables of statewide freight truck flows within, into, and out of Florida for

different geographic resolutions, including the Florida Statewide Model (FLSWM)

traffic analysis zones (TAZs), and

6) Explore the use of ATRI’s GPS data for other applications of interest to Florida,

including the analysis of truck flows out of two sea ports, the re-routing patterns of

trucks after a major highway incident, and the routing patterns of trucks traveling

between Jacksonville and Ocala.

7.2 Project Outcomes and Findings

This section summarizes the outcomes, findings and benefits from the project.

7.2.1 Freight Performance Measures on Florida’s SIS Highway Network

The project resulted in the development of average truck speed data for each (and every) mile of

the SIS highway network for different time periods in the day—AM peak, PM peak, mid-day,

and off peak—using 3 months of ATRI’s truck GPS data in the year 2010. In doing so, it was

found that the existing shape files of the SIS network available with FDOT were either not

accurate enough or they lacked the details (for example, separate links by direction for divided

highways) to derive performance measures using geospatial data. Therefore, a highly-accurate

network was developed to represent highways on the SIS network.

The SIS highway network shape file and the data on average truck speeds by time-of-day

were submitted in a GIS shape file that can be used in an ArcGIS environment to identify the

major freight bottlenecks on Florida’s SIS highway network. In addition to the development of

average speed measures, the project developed example applications of ATRI’s truck GPS data

for measuring truck speed reliability and for highway freight bottleneck analysis.

7.2.2 Algorithms to Convert ATRI’s Raw GPS Data Streams into a Database of Truck

Trips


full potential of the data for freight planning applications. The project resulted in algorithms to

convert the raw GPS data into a data base of truck trips. The results from the algorithms were

subject to different validations to confirm that the algorithms can be used to extract accurate trip

information from the raw GPS data provided by ATRI.

These algorithms were then applied to four months of raw GPS data from ATRI,

comprising a total of 145 million raw GPS data records, to develop a large database of truck trips

traveling within, into, and out of the state. The resulting database comprises more than 1.2

million truck trips traveling within, into, and out of the state. This database of truck trips can be

used for a variety of purposes, including the development of truck travel characteristics and OD

truck flow patterns for different geographical regions in Florida. The database can be used to

calibrate and validate the next-generation statewide freight travel demand model being

developed by Florida Department of Transportation (FDOT). In future work, this database

potentially can be used to develop data on truck trip-chaining and logistics patterns in the state.

86

7.2.3 Analysis of Truck Travel Characteristics in Florida

The truck trip database developed from four months of ATRI’s truck GPS data was used to

analyze a variety of truck travel characteristics in the state of Florida. The truck travel

characteristics analyzed included trip duration, trip length, trip speed, time-of-day profiles, and

origin-destination flows. Each of these characteristics was derived at a statewide level as well as

for different regions in the state—Jacksonville, Tampa Bay, Orlando, Miami, and rest of

Florida—defined based on the freight analysis framework (FAF) zoning system. The time-of-day

profiles of truck trips in all urban regions of Florida show a single peak during the late morning

period as opposed to a bi-modal peak typically observed for passenger travel during morning and

evening peak periods.

In addition, the truck trips were used in conjunction with the GPS data to derive

distributions of OD travel distances, travel times, and travel speeds between over 1,200 TAZ-to-

TAZ OD pairs in the FLSWM. The distributions for each OD pair are reported in the form of

average values and minimum, 5th

percentile, 15th

percentile, 50th

percentile, 85th

percentile, and

maximum values. Comparing the minimum truck travel times measured using GPS data for the

1,200 OD pairs with free-flow travel times used as inputs to FLSWM indicated that the FLSWM

travel times systematically underestimate the truck travel times. A similar comparison with the

travel times extracted from Google Maps also suggested that the Google Maps travel times

underestimate (albeit not as much the FLSWM travel times) the truck travel times. This is most

likely because the travel times used as inputs for the FLSWM and those reported by Google

Maps are predominantly geared toward passenger cars that tend to have higher travel speeds and

better acceleration characteristics. ATRI’s truck GPS data, on the other hand, provide an

opportunity to accurately measure travel times exclusively for trucks (and for different times of

the day).

In addition to the measurement of OD truck travel distances, travel times, and speeds, the

project team performed an exploratory analysis of truck travel routes for more than 1,600 trips

between 10 OD pairs in FLSWM. A preliminary exploratory analysis suggested that a majority

of trips between any OD pair tend to travel largely by similar routes (i.e., the variability in route

choice is not high for the 10 OD pairs examined in this study). Specifically, considerable overlap

was observed among the routes across a large number of trips between an OD pair. Upon further

examination of truck travel routes for a few randomly-sampled OD pairs, it was noticed that

truck travel routes tend to exhibit higher variability for travel over shorter distances within urban

areas than for travel over longer distances. These observations have interesting implications for

future research on understanding and modeling truck route choice. While this study did not delve

further into understanding the route choice patterns of long-haul trucks, this is an important area

for future research using the truck GPS data from ATRI.

7.2.4 Assessment of ATRI’s Truck GPS Data and Its Coverage of Truck Traffic in Florida

This project resulted in a better understanding of ATRI’s truck GPS data in terms of their

coverage of truck traffic in the state of Florida. This includes deriving insights on (a) the types of

trucks (e.g., heavy trucks and medium trucks) present in the data, and (b) the geographical

coverage of the data in Florida, and (c) the proportion of the truck traffic flows in the state

covered by the data.

87

Based on discussions with ATRI and anecdotal evidence, it is known that the major

sources of ATRI data are freight shipping companies that own large trucking fleets, which

typically comprise tractor-trailer combinations (or FHWA vehicle type classes 8–13). However,

a close observation of the data, from following the trucks on Google Earth and examining travel

characteristics of individual trucks, suggested that the data included a small proportion of trucks

that are likely to be smaller trucks that do not necessarily haul freight over long distances. The

project used simple rules to divide the data into two categories: (1) long-haul trucks or heavy

trucks (considered to be FHWA vehicle classification 8–13), and (2) short-haul trucks or medium

trucks. Specifically, trucks that did not make at least one trip of 100-mile length in a two-week

period and those that made more than 5 trips per day were considered “short-haul” trucks. Of a

total of 169,714 unique truck IDs in the data, about 4.6 percent were labeled as short-haul trucks

(or medium trucks) and separated from the remaining long-haul trucks (or heavy trucks). In

future work, it will be useful to derive better definitions of heavy trucks and medium trucks.

Whereas heavy trucks are of primary interest to FLSWM (assuming these are the long-haul

freight carrying trucks), medium trucks are also of potential use for updating the non-freight

truck models. Further, extracting sufficient data on medium trucks potentially can help

understand truck movement within urban regions as well (because a considerable proportion of

truck traffic in urban areas tends to comprise medium trucks).

ATRI’s truck GPS data represented a large sample of truck flows within, coming into,

and going out of Florida. However, the sample was not a census of all trucks traveling in the

state, and it was unknown what proportion of heavy truck flows in the state is represented by

these data. To address this question, truck traffic flows implied by one week of ATRI’s truck

GPS data were compared with truck counts data from more than 200 Telemetered Traffic

Monitoring Sites (TTMS) in the state. The results from this analysis suggest that, at an aggregate

level, the ATRI data provided 10.1 percent coverage of heavy truck flows observed in Florida.

When the coverage was examined separately for different highway facilities (based on functional

classification), the results suggest that the ATRI data provided a representative coverage of truck

flows through different types of highway facilities in the state.

The coverage of ATRI data was examined for different geographical regions in the state

by examining the spatial distribution of the number of truck trips generated at TAZ-level and at

county-level geography. In addition, the percentage of heavy truck traffic covered by ATRI data

at different locations was examined. All these examinations suggested potential geographical

differences in the extent to which ATRI data represent heavy truck traffic volumes at different

locations in the state. For instance, truck trips generated from the Polk County were much higher

than those generated from Hillsborough and Miami-Dade counties. Further, the percentage of

heavy truck traffic covered by ATRI data in the southern part of Florida (within Miami) and the

southern stretch of I-75 is slightly lower compared to the coverage in the northern and central

Florida regions. Such geographical differences (or spatial biases) potentially can be adjusted by

combining ATRI’s truck GPS data with observed data on truck traffic flows at different locations

in the state (from FDOT’s TTMS traffic counting program).

88

7.2.5 Origin-Destination (OD) Tables of Statewide Truck Flows

An important outcome of the project was to use ATRI’s truck GPS data in combination with

other available data to derive OD tables of freight truck flows within, into, and out of the state of

Florida. The OD flow tables were derived at the following levels of geographic resolution:

a) TAZs of the FLSWM, with Florida and the rest of the country divided into about

6,000 TAZs,

b) County-level resolution, in which Florida is represented at a county-level resolution

and the rest of the country is represented at a state-level resolution, and

c) State-level resolution, in which Florida and the rest of the country are represented at a

state-level resolution.

As part of this task, first, the truck trip database developed from four months of ATRI’s

GPS data was converted into OD tables at the TAZ-level spatial resolution used in the FLSWM.

Such an OD table derived only from the ATRI data, however, was not necessarily representative

of the freight truck flows in the state. This was because the ATRI data did not comprise the

census of trucks in the state. Although it was a large sample, it was not necessarily a random

sample and was likely to have spatial biases in its representation of truck flows in the state. To

address these issues, the OD tables derived from the ATRI data were combined with observed

truck traffic volumes at different locations in the state (and outside the state) to derive a more

robust OD table that was representative of the freight truck flows within, into, and out of the

state. To achieve this, a mathematical procedure called origin-destination matrix estimation

(ODME) method was employed to combine the OD flow table generated from the ATRI data

with observed truck traffic volume information at different locations within and outside Florida.

The OD flow table estimated from the ODME procedure was likely to better represent the heavy

truck traffic volumes in the state, as it used the observed truck traffic volumes to weigh the ATRI

data-derived truck OD flow tables.

The truck flow OD tables derived in the project can be used for a variety of different

purposes:

1) To understand the spatial distribution of truck travel demand in the region,

2) To validate, calibrate, and update the heavy truck modeling components of FLSWM,

and

3) Analysis of truck flows into and out of selected locations in Florida.

7.2.6 Explorations of the Use of ATRI’s Truck GPS Data for Other Applications

In addition to the above, this project conducted preliminary explorations of the use of ATRI’s

truck GPS data for the following applications:

a) Analysis of truck flows out of two ports in Florida—Port Blount Island in

Jacksonville and Port Everglades in Fort Lauderdale,

b) Analysis of routing patterns of trucks that used the US 301 roadway to travel between

I-95 around Jacksonville and I-75 around Ocala, and

c) Analysis of changes in truck routing patterns during the closure of a stretch of I-75

near Ocala due to a major multi-vehicle crash in January 2012.

89

These applications were only preliminary explorations conducted as proofs of concept.

Future work can expand on these explorations to conduct full-scale applications.

7.3 Future Work

The work conducted in this project can be extended in several directions of interest to Florida, as

discussed in this section.

7.3.1 Explore the Use of ATRI’s Truck GPS Data for Understanding Urban Freight

Movements and Statewide Non-freight Truck Flows

A significant part of the project was aimed at generating data useful for the FLSWM—for

example, statewide truck OD flows. In future work, it will be useful to explore if ATRI’s truck

GPS data can be used to develop and understand truck flows within urban areas as well. As

mentioned earlier, while the data predominantly comprise heavy trucks that tend to haul freight

over long distances, a non-negligible portion of the data contains medium trucks that tend to

serve local distribution and delivery. Extracting such trucks and analyzing their travel patterns to

understand the extent to which the data covers urban truck flows is a fruitful avenue for future

research. In addition, it will be useful to understand the gaps in these data in terms of what types

of trucks and what industries are not represented in this data. This potentially can help in

augmenting the data with other data sources for use in regional freight travel demand models.

It will be worth exploring the use of these data for generating non-freight truck travel

patterns for FLSWM. Currently, the FLSWM uses Quick-Response Freight Manual (QRFM)

techniques for modeling non-freight truck flows. While QRFM techniques are useful in the

absence of data on non-freight truck flows, it is preferable to develop Florida-specific data to

better model non-freight truck flows in the state.

7.3.2 Estimation of OD Truck Travel Times for FLSWM

This project resulted in measurements of truck travel times (and the distribution of travel times

for different time periods of the day) for more than 1,200 OD pairs in FLSWM. However,

FLSWM has more than 36 million OD pairs (since the model divides Florida and the rest of the

country into more than 6,000 zones). It is not feasible to measure truck travel times using only

the data. However, the travel times measured for 1,200 OD pairs potentially can be used to

develop a travel time estimation model that can be used to estimate the truck travel times for

other OD pairs in FLSWM. Such development of accurate truck travel times that can be input to

the FLSWM is a fruitful avenue for future work.

7.3.3 Analysis of Truck Route Choice

The project explored the route choice patterns of trucks to a limited extent. The data provide a

significant opportunity to better understand truck route choice and to develop truck route choice

models based on accurate data. Combining the data with additional surveys on trucking

companies’ and drivers’ route choice decisions potentially can lead to significant advances in

truck route choice modeling. This can also help improve the truck traffic algorithms currently

used in statewide models.

90

7.3.4 Improvements to ODME

The ODME performed in this study can be improved in different ways. First, the observed truck

traffic volumes used in this study come from FDOT’s telemetric traffic monitoring program (for

more than 200 locations in Florida), Georgia Department of transportation (for several locations

in Georgia), and FHWA’s Vehicle Travel Information System (VTRIS) database (for locations

outside Florida and Georgia). Within the timeframe of this study, the research team could not

gather robust data on observed truck traffic volumes in several southeastern states. For example,

there was little to no traffic count information for states such as Tennessee and a few other

southeastern states. Providing robust truck traffic count data for southeastern states into the

ODME procedure potentially can help in better estimating the truck flows into and out of the

state. Second, the truck GPS data used to develop the seed OD flow table (an input into the

ODME procedure) is Florida-centric. The data do not necessarily provide a reliable picture of the

truck flows between origins and destinations outside Florida. Therefore, using more of ATRI’s

truck GPS data, at least for the southeastern states other than Florida, can potentially help

improve the ODME results. Third, the ODME procedure itself can be improved in different

ways: (a) by allowing different constraints that are specific to different OD pairs, (b) by

exploring the different weighting schemes used to expand the seed matrix, and (c) by improving

the traffic assignment procedure based on observed route choice patterns of trucks. In this

context, analyzing the route choice behavior of trucks is an important avenue for future research

both for improving existing procedures used for traffic assignment and for improving the ODME

procedure for estimating truck OD flows.

7.3.5 Development of Truck Trip Chaining and Logistics Data

This project resulted in procedures for identifying truck stops and truck trips from raw GPS data.

This work can be extended further to derive truck trip chaining and logistics patterns from the

data. In doing so, adding detailed land-use information can help in characterizing the truck travel

patterns.

91

REFERENCES

Bernardin, V. L., Avner J., Short J., Brown L., Nunnally R., & Smith S. (2011). Using large

sample GPS data to develop an improved truck trip table for the Indiana Statewide Model.

Proceedings of 4th Transportation Research Board Conference on Innovations in Travel

Modeling.

Gédéon, C., Florian, M., & Crainic, T. G. (1993). Determining origin-destination matrices and

optimal multiproduct flows for freight transportation over multimodal networks. Transportation

Research Part B: Methodological, 27(5), 351-368.

Giuliano, G., Gordon, P., Pan, Q., Park, J., & Wang, L. (2010). Estimating freight flows for

metropolitan area highway networks using secondary data sources. Networks and Spatial

Economics, 10(1), 73-91.

González-Calderón, C., & Holguín-Veras, J. (2013). Tour-based freight origin-destination

synthesis. Proceedings of Transportation Research Board 92nd Annual Meeting (No. 13-7030).

Holguin-Veras, J., & Patil, G. R. (2007). Integrated origin-destination synthesis model for freight

with commodity-based and empty trip models. Transportation Research Record: Journal of the

Transportation Research Board, 2008(1), 60-66.

Jones, C., Murray, D. C., & Short, J. (2005). Methods of travel time measurement in freight-

significant corridors. American Transportation Research Institute. Available at

http://www.atrionline.org/research/results/Freight%20Performance%20Measures%20TRB%20fo

r%20atri-online.pdf, Accessed on 3/10/2014.

Kuppam, A., Lemp, J., Beagan, D., Livshits, V., Vallabhaneni, L., & Nippani, S. (2014).

Development of a tour-based truck travel demand model using truck GPS data. Proceedings of

Transportation Research Board 93rd Annual Meeting (No. 14-4293).

Ma, X., McCormack, E. D., & Wang, Y. (2011). Processing commercial global positioning

system data to develop a web-based truck performance measures program. Transportation

Research Record: Journal of the Transportation Research Board, 2246(1), 92-100.

Ogden, K. W. (1978). The distribution of truck trips and commodity flow in urban areas: A

gravity model analysis. Transportation Research, 12(2), 131-137.

Short, J. (2010). Bottleneck analysis of 100 freight significant highway locations. American

Transportation Research Institute. Available at http://atri-online.org/2010/05/08/944/, accessed

on 3/10/2014.

Short, J., Pickett, R., & Christianson, J. (2009). Freight performance measures analysis of 30

freight bottlenecks. American Transportation Research Institute. Available at

http://www.mwcog.org/uploads/committee-documents/a15cV1hf20090515125334.pdf, accessed

on 3/10/2014.

http://www.atrionline.org/research/results/Freight%20Performance%20Measures%20TRB%20for%20atri-online.pdf

http://www.atrionline.org/research/results/Freight%20Performance%20Measures%20TRB%20for%20atri-online.pdf

http://atri-online.org/2010/05/08/944/

http://www.mwcog.org/uploads/committee-documents/a15cV1hf20090515125334.pdf

92

Stopher, P. R., Bullock, P., & Jiang, Q. (2003). Visualising trips and travel characteristics from

GPS data. Road and Transport Research, 12(2), 3-14.

Tamin, O. Z., & Willumsen, L. G. (1989). Transport demand model estimation from traffic

counts. Transportation, 16(1), 3-26.

Wardrop, J. G. (1952, June). Road paper. Some theoretical aspects of road traffic research. ICE

Proceedings: Engineering Divisions (Vol. 1, No. 3, pp. 325-362), Thomas Telford.

Zargari, S. A., & Hamedani, S. Y. (2006). Estimation of freight OD matrix using waybill data

and traffic counts in Iran roads. Iranian Journal of Science & Technology, Transaction B,

Engineering, 30(B1).

93

APPENDIX A: EXPLORATORY ANALYSES OF USING ATRI DATA FOR

FREIGHT PLANNING APPLICATIONS IN FLORIDA

In this project, the research team was charged with using the ATRI data for different freight

planning applications of interest to FDOT. Some of the applications, which were intended to be

full-scale applications, are described in the main chapters of the report. Other applications, which

were intended to be of exploratory nature and conducted as proofs-of-concept (not as full

applications) are described in this appendix. The reader will note here that these analyses are

intended to be of exploratory nature, conducted using samples of ATRI’s truck GPS data. Given

the ATRI data sample (and the sample drawn from it) are neither a census of all trucks in Florida

nor a random sample, the results of these analyses must be interpreted with caution.

A.1 I-75 Incident Analysis

ATRI’s truck GPS data can be used to study freight diversion in the event of a road closure or

other significant event. This information is useful for helping to plan for disasters in the future

and also for understanding the impacts of a specific event on a region’s economic activities. As

an example of such an incident, FDOT requested the research team to explore the influence of a

multi-vehicle crash event that occurred on January 29, 2012, which forced the closure of I-75

near Ocala for more than 30 hours. Using ATRI truck GPS data, the research team quantified

unique truck counts on major roadways near the incident area before and during the closure. The

period from 3:00 a.m. on January 22, 2012, to 10:00 a.m. on January 23, 2012 (i.e., one week

before the incident occurred) served as a benchmark to characterize normal traffic flows and

relative volumes experienced on the study area roadways. The study period for the closure due to

the incident extended from 3:00 a.m. on January 29, 2012, to 10:00 a.m. on January 30, 2012. In

the Figure A.1, alternate paths used due to the closure are highlighted by comparing the number

of unique truck counts for each roadway pre-closure to those during the closure. The section that

was closed due to the incident is shown in red. In this example, a color scale depicts loss in

unique truck counts due to diversion in shades of orange. Roadways that were used more

frequently during the closure than pre-closure are shown using a color scale in shades of green.

Such visualizations can be used to understand the truck re-routing patterns during the highway

closure for further detailed analyses of the influence of the incident.

94

Figure A.1 Visualization of Truck Re-routing after Highway (I-75)

Closure due to Multi-vehicle Crash

95

A.2 Trucks Flows from Ports

In another exploratory application, FDOT requested the research team to use the ATRI data to

visualize the travel patterns of trucks leaving two ports in Florida: (1) Jacksonville’s Blount

Island Port and (2) Port Everglades in Fort Lauderdale.

A.2.1 Blount Island Analysis

For this application, ATRI took a sample of trucks that had a nexus with Blount Island in 2010

and followed them from when they left the island for a maximum of 11 hours. Each truck trip,

beginning in Blount Island, was terminated after exhibiting little to no movement for longer than

2 hours. Figure A.2 displays the resulting truck flows from the analysis.

Figure A.2 Analysis of Truck Flows from Blount Island

Once the flows were established, ATRI determined the approximate distribution of route

choices for trucks leaving the island. As Figure A.3 shows, more than 7,500 trips were identified

in the analysis. Of all these trips, a considerable proportion (46.4%) was urban trips that did not

have a destination beyond the I-295 perimeter highway, presumably because there are drayage

trips. Investigating these urban trips further, ATRI discovered that a significant portion of trips

were servicing intermodal facilities in the northern Jacksonville area. As Figure A.4 illustrates,

the vast majority of urban trips (90–95%) did not travel south of I-10 or SR 10.

96

Figure A.3 Truck Trips within 1 Hour of Departing Blount Island

Figure A.4 Urban Truck Trips within 1 Hour of Departing Blount Island

97

A.2.2 Port Everglades Analysis

Similar to the Blount Island analysis, ATRI analyzed truck movements for vehicles servicing

Port Everglades. This analysis isolated a sample of trucks that had a nexus with Port Everglades

in 2010 and followed them from when they left the port area for a maximum of 11 hours. Each

trip was terminated after exhibiting little to no movement for longer than 2 hours. Figure A.5

presents the resulting truck flows generated from the analysis.

Figure A.5 Analysis of Truck Flows from Port Everglades

Figure A.6 shows truck movement within the first hour of leaving Port Everglades. Urban

trips are those that do not leave the Miami–Ft. Lauderdale area (defined by Sawgrass

Expressway on the south, Florida Turnpike on the west, the intracoastal waterway on the east,

and US 441 on the north). As would be expected for a facility so far to the south, the plurality of

trips (49.3%) used I-95 northbound to exit the Miami region. Figure A.7 depicts the 1,341

“urban” trips. Approximately two-thirds of urban trips traveled south of I-595, and one-third

traveled north of I-595.

98

Figure A.6 Truck Trips within 1 Hour of Departing Port Everglades

Figure A.7 Urban Truck Trips within 1 Hour of Departing Port Everglades

99

A.3 I-75 Ocala Analysis

For this task, the research team was requested to characterize truck movements before and after

crossing a traffic counting station on I-75 near Ocala, specifically focusing on the trucks using

US 301. Such analysis can be used to understand the need for a future corridor connecting I-95 in

Jacksonville to I-75 at Ocala. ATRI analyzed one week of truck GPS data (April 26–May 2,

2010) to determine the routes that trucks used before and after crossing the counting station. The

resulting truck flow data were aggregated to determine the share of trips that used specific routes

(Figure A.8). Specifically, Figure A.8 shows where trucks came from before passing Ocala on I-

75. For example, 32.9 percent of the trucks that were on I-75 north of I-10 passed Ocala on I-75

(see northern part of Figure A.8), but only 2.8 percent of the trucks that traveled on I-75 south of

Naples passed Ocala on I-75 (see southern part of Figure A.8).

Figure A.8 Travel Routes of Trucks Crossing the Traffic Counting Station on I-75 at Ocala

FDOT was interested in specifically understanding the degree to which trucks were using

US 301 to travel between I-95 and I-75. To answer this question, ATRI analyzed only trips that

used part of US 301 (Figures A.9 and A.10). For example, in Figure A.9, 13.1 percent of

southbound trips on I-75 (north of Jacksonville) traveled on US 301 to get to I-75 through Ocala.

100

Figure A.9 Routing Patterns of Southbound Trucks Using US 301 in Florida

Figure A.10 Routing Patterns of Northbound Trucks Using US 301 in Florida

101

APPENDIX B: DISTRIBUTIONS OF TRIP LENGTH, TRIP DURATION, TRIP SPEED,

AND TRIP TIME-OF-DAY FOR DIFFERENT SEGMENTS OF TRUCK TRIPS

WITHIN, TO, AND FROM FLORIDA

This appendix provides the distributions of trip length, duration, speed, and time-of-day11

profiles for different segments of the 2.7 million trips derived from four months of ATRI’s truck

GPS data. The different segments include trips starting and ending in different FAF zones in

Florida—Jacksonville, Tampa, Orlando, and Miami. Following are the specific counties in each

of these FAF zones:

Jacksonville FAF zone: Baker, Clay, Duval, Nassau, St. Johns

Miami FAF zone: Broward, Miami-Dade, Palm Beach

Orlando FAF zone: Flagler, Lake, Orange, Sumter, Osceola, Seminole, Volusia

Tampa FAF zone: Hernando, Hillsborough, Pasco, Pinellas

The distributions are provided separately for weekday and weekend trips. Such distributions

potentially can be used for modeling heavy truck trip characteristics within the major regional

models in the state. In addition to the above distributions, for each urbanized county in each of

these FAF zones, the top 10 origins and destinations are provided at the state-level and county-

level geography. In addition, the appendix provides truck trip characteristics for the following

trip segments as well:

a) trips that start and end in Florida (Internal–Internal trips for Florida),

b) trips that start in Florida but end outside Florida (Internal–External trips), and

c) trips that start outside Florida and end in Florida (External–Internal trips).

11

Note: Time-of-day of a trip is determined based on the hour in which the midpoint of the trip falls.

102

Characteristics of Truck Trips within the Tampa FAF Zone

(a)

(b)

Figure B.1 Trip Length Distribution of Trips Starting and Ending in Tampa FAF Zone


0

10

20

30

40

50

60

1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

103

(a)

(b)

Figure B.2 Trip Time Distribution of Trips Starting and Ending in Tampa FAF Zone


0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

104

(a)

(b)

Figure B.3 Trip Speed Distribution of Trips Starting and Ending in Tampa FAF Zone


0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps



Average Trip Speed

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps



Average Trip Speed

105

(a)

(b)

Figure B.4 Time-of-Day Profile of Trips Starting and Ending in Tampa FAF Zone


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

106

Figure B.5 Destinations of Trips Starting in Hillsborough County (87,701 Trips)


other states

Georgia

Alabama

North Carolina

South Carolina

Tennessee

Top 10 destinations other

than Hillsborough County

Polk

Pinellas

Orange

Manatee

Duval

Pasco

Lee

Sarasota

Lake

Miami-Dade

107

Figure B.6 Origins of Trips Ending in Hillsborough County (86,254 Trips)

Top 10 origins other than

Hillsborough County

Polk

Pinellas

Orange

Pasco

Manatee

Duval

Lee

Sarasota

Osceola

Miami-Dade


states

Georgia

Alabama

North Carolina

South Carolina

Tennessee

108

Characteristics of Truck Trips within the Orlando FAF Zone

(a)

(b)

Figure B.7 Trip Length Distribution of Trips Starting and Ending in Orlando FAF Zone


0

10

20

30

40

50

60

1-1

0

10-2

0

20-3

0

30-4

0

40-5

0

50-6

0

60-7

0

70-8

0

80-9

0

90-1

00

100

-11

0

110

-12

0

120

-13

0

130

-14

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

1-1

0

10-2

0

20-3

0

30-4

0

40-5

0

50-6

0

60-7

0

70-8

0

80-9

0

90-1

00

100

-11

0

110

-12

0

120

-13

0

130

-14

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

109

(a)

(b)

Figure B.8 Trip Time Distribution of Trips Starting and Ending in Orlando FAF Zone


0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

110

(a)

(b)

Figure B.9 Trip Speed Distribution of Trips Starting and Ending in Orlando FAF Zone


0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

111

(a)

(b)

Figure B.10 Time-of-Day Profile of Trips Starting and Ending in Orlando FAF Zone


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

112

Figure B.11 Destinations of Trips Starting in Orange County (94,575 Trips)


other states

Georgia

Alabama

South Carolina

North Carolina

Texas


Orange County

Polk

Duval

Lake

Osceola

Seminole

Hillsborough

Volusia

Brevard

Miami-Dade

Broward

113

Figure B.12 Origins of Trips Ending in Orange County (92,994 Trips)

Top 10 origins other than

Orange County

Polk

Hillsborough

Duval

Seminole

Osceola

Lake

Miami-Dade

Volusia

Brevard

Alachua

Top 5 origins among

other states

Georgia

Alabama

North Carolina

South Carolina

Texas

114

Characteristics of Truck Trips within the Jacksonville FAF Zone

(a)

(b)

Figure B.13 Trip Length Distribution of Trips Starting and Ending in Jacksonville FAF

Zone during (a) Weekdays (56,319 trips), and (b) Weekend (5,789 trips)

0

10

20

30

40

50

60

1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

115

(a)

(b)

Figure B.14 Trip Time Distribution of Trips Starting and Ending in Jacksonville FAF Zone


0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

116

(a)

(b)

Figure B.15 Trip Speed Distribution of Trips Starting and Ending in Jacksonville FAF

Zone during (a) Weekdays (56,319 trips), and (b) Weekend (5,789 trips)

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

117

(a)

(b)

Figure B.16 Time-of-Day Profile of Trips Starting and Ending in Jacksonville FAF Zone


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tip

s


Trip Start Time

Trip End Time

118

Figure B.17 Destinations of Trips Starting in Duval County (110,314 Trips)


other states

Georgia

North Carolina

South Carolina

Alabama

Tennessee

Top 10 destinations other than Duval

County

Orange

Nassau

Polk

Putnam

Miami-Dade

Hillsborough

Broward

Alachua

Clay

Marion

119

Figure B.18 Origins of Trips Ending in Duval County (111,023 Trips)

Top 10 origins other than Duval County

Orange

Polk

Nassau

Hillsborough

Putnam

Miami-Dade

Alachua

Palm Beach

Baker

Clay


states

Georgia

South Carolina

North Carolina

Alabama

Texas

120

Characteristics of Truck Trips within the Miami FAF Zone

(a)

(b)

Figure B.19 Trip Length Distribution of Trips Starting and Ending in Miami FAF Zone


0

10

20

30

40

50

60

1-1

0

10-2

0

20-3

0

30-4

0

40-5

0

50-6

0

60-7

0

70-8

0

80-9

0

90-1

00

100

-11

0

110

-12

0

120

-13

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

1-1

0

10-2

0

20-3

0

30-4

0

40-5

0

50-6

0

60-7

0

70-8

0

80-9

0

90-1

00

100

-11

0

110

-12

0

120

-13

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

121

(a)

(b)

Figure B.20 Trip Time Distribution of Trips Starting and Ending in Miami FAF Zone


0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

122

(a)

(b)

Figure B.21 Trip Speed Distribution of Trips Starting and Ending in Miami FAF Zone


0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

123

(a)

(b)

Figure B.22 Time-of-Day Profile of Trips Starting and Ending in Miami FAF Zone


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

124

Figure B.23 Destinations of Trips Starting in Miami-Dade County (70,273 Trips)


other states

Georgia

South Carolina

Alabama

Texas

North Carolina

Top 10 destinations other than Miami-

Dade County

Broward

Palm Beach

Polk

Duval

Orange

Hillsborough

Lake

De Soto

Putnam

St. Lucie

125

Figure B.24 Origins of Trips Ending in Miami-Dade County (69,274 Trips)

Top 10 origins other than Miami-Dade

County

Broward

Polk

Palm Beach

Duval

Orange

Hillsborough

De Soto

Osceola

Seminole

Putnam


states

Georgia

Texas

Tennessee

North Carolina

Alabama

126

Characteristics of Truck Trips Starting and Ending in the Remainder of Florida

(a)

(b)

Figure B.25 Trip Length Distribution of Trips Starting and Ending in Other Regions in

Florida during Weekdays (527,484 Trips), and (b) Weekend (73,678 Trips)

0

10

20

30

40

50

60

1-50 50-100 100-200 200-500 500-800

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip length (miles)

0

10

20

30

40

50

60

1-50 50-100 100-200 200-500 500-800

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

127

(a)

(b)

Figure B.26 Trip Time Distribution of Trips Starting and Ending in Other Regions in


0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

30

35

40

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

> 2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

128

(a)

(b)

Figure B.27 Trip Speed Distribution of Trips Starting and Ending in Other Regions in


0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

129

(a)

(b)

Figure B.28 Time-of-Day Profile of Trips Starting and Ending in Other Regions in Florida

during (a) Weekdays (527,484 Trips), and (b) Weekend (73,678 Trips)

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

130

Figure B.29 Destinations of Trips Starting in Polk County (133,365 Trips)


other states

Georgia

Alabama

South Carolina

Texas

North Carolina

Top 10 destinations other than Polk

County

Hillsborough

Orange

Lake

Osceola

Duval

Miami-Dade

De Soto

Broward

Pinellas

Manatee

131

Figure B.30 Origins of Trips Ending in Polk County (132,393 Trips)

Top 10 origins other than Polk County

Orange

Hillsborough

Lake

Osceola

Miami-Dade

Duval

De Soto

Pinellas

Broward

Manatee


states

Georgia

Alabama

North Carolina

South Carolina

Texas

132

Characteristics of Truck Trips Starting and Ending in Florida (I-I Trips for Florida)

(a)

(b)

Figure B.31 Trip Length Distribution of Trips Starting and Ending in Florida


0

10

20

30

40

50

60

70

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

70

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

133

(a)

(b)

Figure B.32 Trip Time Distribution of Trips Starting and Ending in Florida


0

5

10

15

20

25

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

134

(a)

(b)

Figure B.33 Trip Speed Distribution of Trips Starting and Ending in Florida


0

5

10

15

20

25

30

35

40

45

50

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

45

50

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

135

(a)

(b)

Figure B.34 Time-of-Day Profile of Trips Starting and Ending in Florida


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

136

Characteristics of Truck Trips Starting in Florida and Ending outside Florida (I-E Trips)

(a)

(b)

Figure B.35 Trip Length Distribution of Trips Starting in Florida and Ending outside

Florida during (a) Weekdays (138,816 trips), and (b) Weekend (13,900 trips)

0

10

20

30

40

50

60

70

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

100

0-2

000

200

0-3

000

>3

00

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

70

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

100

0-2

000

200

0-3

000

>3

00

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

137

(a)

(b)

Figure B.36 Trip Time Distribution of Trips Starting in Florida and Ending outside


0

5

10

15

20

25

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

138

(a)

(b)

Figure B.37 Trip Speed Distribution of Trips Starting in Florida and Ending outside


0

5

10

15

20

25

30

35

40

45

50

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

45

50

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

139

(a)

(b)

Figure B.38 Time-of-Day Profile of Trips Starting in Florida and Ending outside Florida


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps

Time of Day (0 is 12AM)

Trip Start Time

Trip End Time

140

Figure B.39 Destinations of Trips Starting in Florida (1,102,873 Trips)

Top 10 destinations other than Florida

Georgia

Alabama

South Carolina

North Carolina

Mississippi

Texas

Tennessee

Louisiana

Virginia

Pennsylvania

141

Characteristics of Truck Trips Starting outside Florida and Ending in Florida (E-I Trips)

(a)

(b)

Figure B.40 Trip Length Distribution of Trips Starting outside Florida and Ending in


0

10

20

30

40

50

60

70

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

100

0-2

000

200

0-3

000

>3

00

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

0

10

20

30

40

50

60

70

1-5

0

50-1

00

100

-20

0

200

-50

0

500

-10

00

100

0-2

000

200

0-3

000

>3

00

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Length (miles)

142

(a)

(b)

Figure B.41 Trip Time Distribution of Trips Starting outside Florida and Ending in


0

5

10

15

20

25

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

0

5

10

15

20

25

0-5

5-1

0

10-1

5

15-3

0

30-6

0

60-9

0

90-1

20

120

-18

0

180

-24

0

240

-30

0

300

-42

0

420

-54

0

540

-72

0

720

-14

40

144

0-2

880

>2

88

0

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Time (minutes)

Trip Time in Motion

Total Trip Time

143

(a)

(b)

Figure B.42 Trip Speed Distribution of Trips Starting outside Florida and Ending in


0

5

10

15

20

25

30

35

40

45

50

0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

0

5

10

15

20

25

30

35

40

45

50

0-1

0

10-2

0

20-3

0

30-4

0

40-5

0

50-6

0

60-7

0

70-8

0

80-9

0

90-1

00

Per

cen

tag

e o

f T

ruck

Tri

ps

Trip Speed (mph)


Average Trip Speed

144

(a)

(b)

Figure B.43 Time-of-Day Profile of Trips Starting outside Florida and Ending in Florida


0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

0

1

2

3

4

5

6

7

8

9

10

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

Per

cen

tag

e o

f T

ruck

Tri

ps


Trip Start Time

Trip End Time

145

Figure B.44 Origins of Trips Ending in Florida (1,097,614 Trips)

Top 10 origins other than Florida

Georgia

Alabama

South Carolina

North Carolina

Texas

Tennessee

Mississippi

Louisiana

Pennsylvania

Virginia

146

APPENDIX C: TRAVEL ROUTES AND TRAVEL TIME MEASUREMENTS

FOR 10 OD PAIRS

This appendix provides maps of travel routes and histograms of travel time distributions between

10 TAZ-to-TAZ OD pairs in FLSWM. The 10 OD pairs were selected strategically to include a

randomly-selected sample of 2 OD pairs in each of the following five travel distance bands:

within 100 miles, 100–200 miles, 200–500 miles, 500–1000 miles, and above 1,000 miles. For

each of these OD pairs, the route choices of all trips extracted from the data are depicted as on-

route GPS coordinates. Specifically, Figures D.1 through D.10 show maps with the route

choices, one figure for each OD pair. These maps can be used to gain a preliminary

understanding of the route choice patterns of long-haul trucks.

In addition to route choice maps, the appendix provides histograms of OD travel time

distributions for the same 10 OD pairs. Specifically, the distributions for three different types of

travel times are provided: (1) total travel time, which measures the time between trip start and

tripe end, including the time spent at all intermediate stops such as rest stops, traffic congestion

stops, and fueling stops; (2) travel time in motion including non-significant stops, which

measures the travel time excluding significant stops such as rest stops but includes non-

significant stops such as traffic congestion stops and fueling stops; and (3) travel time in motion

excluding non-significant stops, which excludes the time spent all intermediate stops, including

rest stops, traffic congestion stops and fueling stops. These distributions are provided for

different time periods of the day such as AM peak, mid-day, PM peak, and night-time. This

information can be used to validate the freight component of the new FLSWM recently

developed by FDOT (by comparing the measured travel times provided in this appendix with the

travel times estimated or output from the model).

147

Figure C.1 Route Choice for 365 Trips for OD Pair “3124-3703”


Destination TAZ#1863 Polk County, FL

Origin TAZ# 4526 Hillsborough County, FL

Destination TAZ#3703

Polk County, FL

Origin TAZ#3124

Orange County, FL

148

Figure C.3 Route Choice for 327 Trips between OD Pair “2228-4227”


Origin TAZ# 2228

Alachua County, FL

Destination TAZ# 4227

Orange County, FL

Origin TAZ# 3662 Manatee County, FL

Destination TAZ#3124 Orange County, FL

149



Destination TAZ#792 Escambia County, FL

Origin TAZ# 5983 Jefferson County, Alabama

Origin TAZ #2420,

Polk County, FL

Destination TAZ # 4147,

Miami-Dade County, FL

150


Figure C.8 Route Choice 38 Trips between OD Pair “5073-6117”

Destination TAZ#5819 Gordon County, Georgia

Origin TAZ# 4035 Palm Beach County, FL

Destination TAZ # 6117

Davidson, Tennessee

Origin TAZ # 5073

Duval County, Florida

151


Figure C.10 Route Choice for 28 Trips between OD Pair “2355-6176”

Destination TAZ#6086 Texas

Origin TAZ# 413 Duval County, FL

Destination TAZ# 6176 Massachusetts

Origin TAZ# 2355 Lake County, FL

152

Origin TAZ: 3124, Destination TAZ: 3703

Figure C.11 Travel Time Distribution for Trips between OD Pair “3124-3703” (384 Trips)

Figure C.12 Trips between OD Pair “3124-3703” during Weekdays (284 Trips)

0

10

20

30

40

50

60

70

80

90

51-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

0Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time

Trip time in motion including non-significant stops

Trip time in motion minus non-significant stops

0

10

20

30

40

50

60

70

80

90

51-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

0Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



153

Figure C.13 Mid-day Trips between OD Pair “3124-3703” during Weekdays (111 Trips)

Figure C.14 PM Peak Trips between OD Pair “3124-3703” during Weekdays (88 Trips)

Note: In this case, total trip time and trip time in motion including non-significant stops are equal since

the trucks did not make any significant stops in rest areas/gas stations.

0

10

20

30

40

50

60

70

80

90

51-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

0Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

51-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

0Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time


154

Figure C.15 Night-time Trips between OD Pair “3124-3703” during Weekdays (83 Trips)

Note: In this case, the total trip time and trip time in motion including non-significant stops are equal

since the trucks did not make any significant stops in rest areas/gas stations. Therefore, a separate

histogram is not provided for trip time in motion including non-significant stops.

0

10

20

30

40

50

60

70

80

90

51-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

0Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time


155




0

10

20

30

40

50

60

70

80

90

33-4

5

45-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

5

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

33-4

5

45-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

5

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



156


0

10

20

30

40

50

60

70

80

90

33-4

5

45-6

0

60-7

5

75-9

0

90-1

05

105

-12

0

120

-13

5

135

-15

0

150

-16

5

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



157


Figure C.19 Travel Time Distribution for Trips between OD Pair “2228-4227”

during Weekdays (341 Trips)


0

10

20

30

40

50

60

70

80

90

94-1

20

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-30

0

300

-33

0

330

-35

7

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

94-1

20

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-30

0

300

-33

0

330

-35

7

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



158

Figure C.21 AM Peak Trips between OD Pair “2228-4227” during Weekdays (173 Trips)

0

10

20

30

40

50

60

70

80

90

94-1

20

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-30

0

300

-33

0

330

-35

7

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



159




0

10

20

30

40

50

60

70

80

90

111

-12

0

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-29

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

111

-12

0

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-29

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



160



0

10

20

30

40

50

60

70

80

90

111

-12

0

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-29

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

111

-12

0

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-29

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



161


0

10

20

30

40

50

60

70

80

90

111

-12

0

120

-15

0

150

-18

0

180

-21

0

210

-24

0

240

-27

0

270

-29

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



162




0

10

20

30

40

50

60

70

80

90

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-30

0

300

-36

0

360

-42

0

420

-54

0

540

-66

0

660

-78

0

780

-90

0

900

-98

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-30

0

300

-36

0

360

-42

0

420

-54

0

540

-66

0

660

-78

0

780

-90

0

900

-98

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



163



0

10

20

30

40

50

60

70

80

90

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-30

0

300

-36

0

360

-42

0

420

-54

0

540

-66

0

660

-78

0

780

-90

0

900

-98

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-30

0

300

-36

0

360

-42

0

420

-54

0

540

-66

0

660

-78

0

780

-90

0

900

-98

3

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



164

Origin TAZ: 2420, Destination TAZ: 4147 12



12 In this case, total trip time and trip time in motion including non-significant stops are equal since the trucks did

not make any significant stops in rest areas/gas stations.

0

10

20

30

40

50

60

70

80

90

203

-21

0

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-28

5

285

-30

0

300

-31

5

315

-33

5Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (Minutes)

Total trip time


0

10

20

30

40

50

60

70

80

90

203

-21

0

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-28

5

285

-30

0

300

-31

5

315

-33

5Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time


165


0

10

20

30

40

50

60

70

80

90

203

-21

0

210

-22

5

225

-24

0

240

-25

5

255

-27

0

270

-28

5

285

-30

0

300

-31

5

315

-33

5Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

s

Trip Time (minutes)

Total trip time


166





0

10

20

30

40

50

60

70

80

90

541

-57

0

570

-60

0

600

-63

0

630

-66

0

660

-72

0

720

-78

0

780

-90

0

900

-11

40

114

0-

126

0

126

0-

143

0

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

541

-57

0

570

-60

0

600

-63

0

630

-66

0

660

-72

0

720

-78

0

780

-90

0

900

-11

40

114

0-1

260

126

0-1

430Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



167




0

10

20

30

40

50

60

70

80

904

68

-48

0

480

-54

0

540

-60

0

600

-66

0

660

-72

0

720

-78

0

780

-84

0

840

-90

0

900

-12

00

120

0-1

500

150

0-1

800

180

0-5

744P

erce

nta

ge

of

Tru

ck T

rip

s b

etw

een

OD

Pa

ir

Trip Time ( Minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

468

-48

0

480

-54

0

540

-60

0

600

-66

0

660

-72

0

720

-78

0

780

-84

0

840

-90

0

900

-12

00

120

0-1

500

150

0-1

800

180

0-5

744

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



168


0

10

20

30

40

50

60

70

80

90

468

-48

0

480

-54

0

540

-60

0

600

-66

0

660

-72

0

720

-78

0

780

-84

0

840

-90

0

900

-12

00

120

0-1

500

150

0-1

800

180

0-5

744P

erce

nta

ge

of

Tru

ck T

rip

s b

etw

een

OD

Pa

ir

Trip Time (minutes)

Total trip time



169





0

10

20

30

40

50

60

70

80

90

946

-10

20

102

0-1

080

108

0-1

200

120

0-1

320

132

0-1

440

144

0-1

560

156

0-1

800

180

0-2

040

204

0-2

760

276

0-3

480

348

0-5

640

564

0-5

686Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

946

-10

20

102

0-1

080

108

0-1

200

120

0-1

320

132

0-1

440

144

0-1

560

156

0-1

800

180

0-2

040

204

0-2

760

276

0-3

480

348

0-5

640

564

0-5

686Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



170



during Weekday (28 Trips)


0

10

20

30

40

50

60

70

80

90

126

4-1

320

132

0-1

380

138

0-1

440

144

0-1

500

150

0-1

560

156

0-1

620

162

0-2

040

204

0-2

160

216

0-2

280

228

0-2

400

240

0-2

640

264

0-2

880

288

0-3

753

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



0

10

20

30

40

50

60

70

80

90

126

4-1

320

132

0-1

380

138

0-1

440

144

0-1

500

150

0-1

560

156

0-1

620

162

0-2

040

204

0-2

160

216

0-2

280

228

0-2

400

240

0-2

640

264

0-2

880

288

0-3

753

Per

cen

tag

e o

f T

ruck

Tri

ps

bet

wee

n O

D P

air

Trip Time (minutes)

Total trip time



171

APPENDIX D: SEASONALITY ANALYSIS

As mentioned in Chapter 2 of this report, since ATRI data consist of tractor-trailer trucks of

classes 8 and above, seasonality analysis was performed for tractor-trailer trucks and single-unit

trucks separately. The tables and charts in this appendix represent the analysis of truck counts

during different months of year 2010 as well as the aggregated counts for different seasons based

on Table D.1, which shows the dates forming each season. Also, analysis was conducted for

different facility types as well as area types during different time periods. The truck counts were

obtained from Florida Department of Transportation for 252 locations across Florida.

Table D.1 Dates Forming Each Season

Season Starting Date Ending Date

Summer 21-Jun 22-Sep

Autumn 23-Sep 20-Dec

Winter 21-Dec 20-Mar

Spring 21-Mar 20-Jun

172

D.1 Single-Unit Trucks

Table D.2 Seasonality Analysis of Single-Unit Trucks by Facility Type

FTYPE Total

Sites AADT Summer Autumn Winter Spring Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

11 8 3359 2895 2985 3527 3240 3359 3667 3252 3288 3270 3285 3120 2958 2815 2833 2918 2816

12 40 2066 2004 2032 1989 2089 1790 1980 2145 2007 1965 2091 2011 1975 1958 2102 1981 1893

16 2 2939 2885 2941 2903 2957 2438 3081 3013 3006 2913 3007 2902 2888 2868 3036 2823 2747

21 85 849 833 871 865 893 833 860 890 914 874 861 815 848 864 908 854 821

23 5 891 882 836 1030 880 861 1049 1049 909 840 893 887 870 889 844 812 810

24 1 619 637 678 575 592 552 568 592 582 583 631 641 645 649 692 642 674

31 77 313 302 309 333 337 323 340 340 335 337 326 286 295 299 310 299 297

33 1 345 336 357 348 351 327 377 342 367 347 337 337 317 363 329 369 324

41 2 364 341 366 406 368 395 406 395 402 356 323 295 371 372 375 341 349

46 7 334 327 325 343 352 329 348 365 352 340 354 334 322 323 335 312 304

83 4 4894 4871 4945 4751 4920 4618 4832 4771 4244 4121 4996 4828 4996 4738 5550 4892 5045

91 9 2697 2637 2749 2541 2785 2383 2627 2648 2829 2338 2855 2658 2565 2550 2667 2566 2563

92 10 1306 1280 1242 1409 1343 1353 1390 1338 1448 1342 1349 1293 1283 1240 1330 1274 1194

93 1 1726 1751 1691 1685 1827 1611 1720 1769 1794 1801 1940 1765 1704 1720 1721 1638 1578

Table D.3 Description of Facility Types (FTYPE)

Facility Type Description Facility Type Description

11 Urban Freeway Group 1 (cities of 500,000 or more) 33 Undivided Arterial Class 1b with Turn Bays

12 Other Freeway (not in Group 1) 41 Major Local Divided Roadway

16 Controlled Access Expressway 46 Other Local Undivided Roadway without Turn Bays

21 Divided Arterial Unsignalized (55 mph) 83 Freeway Group 1 HOV Lane (Non-Separated)

23 Divided Arterial Class 1a 91 Freeway Group 1 Toll Facility

24 Divided Arterial Class 1b 92 Other Freeway Toll Facility

31 Undivided Arterial Unsignalized with Turn Bays 93 Expressway/Parkway Toll Facility

173

Figure D.1 Florida Truck Flows (Counts for Single-Unit Trucks) for

Facility Type: Urban Freeway

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

AADT Spring Summer Autumn Winter

Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

174


Facility Type: Other Freeway

0

500

1000

1500

2000

2500

3000

3500

4000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

175


Facility Type: Controlled-Access Expressway

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000


Tru

cks

pe

r D

ay

Season

Mean 95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

176


Facility Type: Divided Arterial Unsignalized

0

200

400

600

800

1000

1200

1400

1600

1800

2000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

200

400

600

800

1000

1200

1400

1600

1800

2000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

177


Facility Type: Divided Arterial Class 1a

0

200

400

600

800

1000

1200

1400

1600

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

200

400

600

800

1000

1200

1400

1600


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

178


Facility Type: Undivided Arterial Unsignalized with Turn Bays

0

100

200

300

400

500

600

700

800

900

1000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

100

200

300

400

500

600

700

800

900

1000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

179


Facility Type: Major Local Divided Roadway

0

100

200

300

400

500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

50

100

150

200

250

300

350

400

450

500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

180


Facility Type: Other Local Undivided Roadway without Turn Bays

0

100

200

300

400

500

600

700

800

900

1000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

100

200

300

400

500

600

700

800

900

1000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

181


Facility Type: Freeway Group 1 HOV Lane (Non-Separated)

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

1000

2000

3000

4000

5000

6000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

182


Facility Type: Freeway Group 1 Toll Facility

0

500

1000

1500

2000

2500

3000

3500

4000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

183

Figure D.11 Florida Truck Flows (Counts for Single Unit-Trucks) for

Facility Type: Other Freeway Toll Facility

0

500

1000

1500

2000

2500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

184

Table D.4 Seasonality Analysis of Single-Unit Trucks by Area Type

ATYPE Total


12 1 243 262 228 234 253 224 240 235 239 263 282 274 256 249 227 222 209

14 1 619 637 678 575 592 552 568 592 582 583 631 641 645 649 692 642 674

21 3 1483 1478 1526 1327 1512 1275 958 1523 1542 1456 1500 1464 1501 1483 1597 1462 1415

31 53 1513 1478 1537 1482 1554 1329 1399 1524 1597 1497 1546 1458 1466 1491 1580 1461 1435

32 6 1692 1683 1671 1686 1820 1583 1709 1819 1835 1766 1839 1750 1675 1640 1684 1631 1581

33 23 1377 1283 1331 1382 1244 1277 1420 1434 1297 1249 1246 1192 1205 1368 1435 1337 1293

34 1 262 0 284 255 0 253 262 0 0 0 0 0 0 0 0 0 262

41 1 1214 1277 0 1120 1231 1081 1098 1195 1257 1188 1309 1333 1237 1206 0 0 0

42 36 1840 1728 1830 1746 1810 1687 1836 1740 1742 1645 1756 1674 1763 1757 1772 1922 1782

43 1 197 215 153 184 237 165 179 242 238 214 264 256 193 181 164 137 139

51 17 1099 1070 1108 1128 1113 1068 1164 1155 1140 1077 1108 1066 1075 1055 1127 1093 1046

52 109 595 598 596 598 648 571 600 609 623 577 626 595 598 578 618 593 554

Table D.5 Description of Area Types (ATYPE) Area Type Description

12 Urbanized Area (under 500,000) Primary City Central Business District

14 Non-Urbanized Area Small City Downtown

21 All Central Business District

31 Residential Area of Urbanized Areas

32 Undeveloped Portions of Urbanized Areas

33 Transitioning Areas/Urban Areas over 5,000 Population

34 Beach Residential (per SERPM)

41 High Density Outlying Business District

42 Other Outlying Business District

43 Beach OBD (per SERPM)

51 Developed Rural Areas/Small Cities under 5,000 Population

52 Undeveloped Rural Areas

185


Area Type: Central Business District

0

500

1000

1500

2000

2500

3000

3500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

186


Area Type: Residential Area of Urbanized Areas

0

500

1000

1500

2000

2500

3000

3500

4000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

187


Area Type: Undeveloped Portions of Urbanized Areas

0

500

1000

1500

2000

2500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

188


Area Type: Transitioning Areas/Urban Areas over 5,000 Population

0

500

1000

1500

2000

2500

3000

3500

4000

4500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

189


Area Type: Other Outlying Business District

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

190


Area Type: Developed Rural Areas/Small Cities under 5,000 Population

0

500

1000

1500

2000

2500

3000

3500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

191


Area Type: Undeveloped Rural Areas

0

500

1000

1500

2000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

192

D.2 Tractor-Trailer Trucks

Table D.6 Seasonality Analysis of Tractor-Trailer Trucks by Facility Type

FTYPE Total


11 8 4670 4194 4360 5262 5237 5065 5294 5258 5217 5958 5691 5265 4238 4144 4213 4272 4195

12 40 6129 5957 6147 5929 6288 5762 6224 6390 6281 5995 6138 5824 5965 5906 6214 5860 5646

16 2 2477 2474 2565 2318 2453 1883 2465 2502 2482 2390 2545 2502 2506 2417 2587 2524 2395

21 85 748 752 741 756 831 730 767 817 860 833 757 757 747 766 793 754 713

23 5 1000 969 1032 1349 1086 1081 1354 1425 1123 1080 1033 971 969 945 842 1017 945

24 1 374 406 362 320 423 301 318 368 440 404 432 413 398 399 388 348 317

31 77 319 307 304 299 354 288 294 322 363 359 329 306 302 298 309 302 290

33 1 223 242 214 187 246 176 195 216 247 246 260 256 235 221 228 211 179

41 2 561 576 549 536 591 519 543 558 599 592 603 573 588 544 573 543 500

46 7 782 768 762 784 829 743 781 867 843 818 811 768 766 754 777 737 713

83 4 6883 6690 7039 6848 6886 6704 6941 6964 6104 5779 6865 6624 6746 6632 7929 7023 7594

91 9 2918 2872 2983 2998 3301 2799 3066 3179 3355 3019 3320 3118 2804 2807 2898 2840 2855

92 10 1873 1845 1816 1640 1930 1571 1157 1434 1853 1862 1957 1830 1843 1829 1920 1865 1731

93 1 1255 1217 1239 1227 1368 1176 1246 1315 1392 1350 1320 1225 1197 1217 1256 1215 1170

193

Figure D.19 Florida Truck Flows (Counts for Tractor-Trailer Trucks) for

Facility Type: Urban Freeway Group 1

0

1000

2000

3000

4000

5000

6000

7000

8000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

1000

2000

3000

4000

5000

6000

7000

8000


Tru

cks

pe

r D

ay

Season

Mean 95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

194


Facility Type: Other Freeway

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

195


Facility Type: Controlled Access Expressway

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

196


Facility Type: Divided Arterial Unsignalized

0

500

1000

1500

2000

2500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

197


Facility Type: Divided Arterial Class 1a

0

500

1000

1500

2000

2500

3000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

198


Facility Type: Undivided Arterial Unsignalized with Turn Bays

0

100

200

300

400

500

600

700

800

900

1000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

100

200

300

400

500

600

700

800

900

1000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

199


Facility Type: Major Local Divided Roadway

0

100

200

300

400

500

600

700

800

900

1000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

100

200

300

400

500

600

700

800

900


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

200


Facility Type: Other Local Undivided Roadway without Turn Bays

0

500

1000

1500

2000

2500

3000

3500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

201


Facility Type: Freeway Group 1 HOV Lane

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

1000

2000

3000

4000

5000

6000

7000

8000

9000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

202


Facility Type: Freeway Group 1 Toll Facility

0

500

1000

1500

2000

2500

3000

3500

4000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

203


Facility Type: Other Freeway Toll Facility

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r d

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

204

Table D.7 Seasonality Analysis of Tractor-Trailer Trucks by Area Type

ATYPE Total Sites AADT Summer Autumn Winter Spring Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

12 1 7016 6561 7119 7019 7518 6675 7097 7491 7613 7529 7120 6509 6560 6505 7035 7148 6908

14 1 8384 8117 8553 8355 8751 8023 8405 8857 8664 8805 8586 8012 8226 8073 8547 8337 8150

21 3 2069 2115 2118 1859 1982 1778 1906 1933 1980 1257 1261 1177 2153 2077 1806 2032 2114

31 53 5140 4842 5092 5011 5109 4814 5271 5419 4905 4754 4767 4603 4737 4893 5250 4864 4555

32 6 630 639 584 839 1001 814 606 665 1037 948 977 950 643 613 618 564 525

33 23 522 486 453 535 594 501 473 523 624 533 540 498 507 496 486 453 425

34 1 287 282 283 282 307 275 288 289 303 313 306 283 277 272 292 290 259

41 1 1057 1068 1003 1064 1121 1042 1094 1068 1135 1074 1169 1034 1088 1071 1097 934 897

42 36 872 899 917 875 945 838 900 932 949 1022 874 928 901 912 973 1014 975

43 1 1188 1217 1169 1141 1248 1095 1157 1211 1255 1204 1320 1186 1214 1198 1209 1117 1088

51 17 767 745 676 764 773 683 5647 800 275 736 755 706 582 717 692 661 627

52 109 961 972 985 865 1048 853 853 909 1012 990 1053 991 1007 968 961 982 912

205


Area Type: Central Business District

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

6500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

6500


Tru

cks

pe

r D

ay

Season

Mean95th

Percentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

206


Area Type: Residential Areas of Urbanized Areas

0

500

1000

1500

2000

2500

3000

3500

4000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000


Tru

cks

pe

r D

ay

Season

Mean 95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

207


Area Type: Undeveloped Portions of Urbanized Areas

0

500

1000

1500

2000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

208


Area Type: Transitioning Areas/Urban Areas over 5,000 Population

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

1000

2000

3000

4000

5000

6000

7000

8000

9000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

209


Area Type: Other Outlying Business District

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

6500

7000

7500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

6500

7000

7500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

210

Figure D. 35 Florida Truck Flows (Counts for Tractor-Trailer Trucks) for

Area Type: Developed Rural Areas/Small Cities

0

500

1000

1500

2000

2500

3000

3500

4000

4500

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

500

1000

1500

2000

2500

3000

3500

4000

4500


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

211


Area Type: Undeveloped Rural Areas

0

1000

2000

3000

4000

5000

6000

7000

AA

DT

Jan

uar

y

Feb

ruar

y

Mar

ch

Ap

ril

May

Jun

e

July

Au

gust

Sep

tem

be

r

Oct

ob

er

No

vem

be

r

De

cem

ber

Tru

cks

pe

r D

ay

Month

Mean

0

1000

2000

3000

4000

5000

6000

7000


Tru

cks

pe

r D

ay

Season

Mean

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

95thPercentile

75thPercentile

Median

25thPercentile

5thPercentile

212

(a)

(b)

Figure D.37 a) Distribution of Daily Truck Counts in Traffic Sites;

b) Same Distribution by Facility Type

0

5

10

15

20

25

13-1

00

100

-20

0

200

-30

0

300

-40

0

400

-50

0

500

-10

00

100

0-2

000

200

0-3

000

300

0-4

000

400

0-5

000

500

0-6

282

Per

cen

tag

e o

f T

TM

S S

ites

Number of Daily Truck Counts (Classes 8-13)

0

10

20

30

40

50

60

70

13-1

00

100

-20

0

200

-30

0

300

-40

0

400

-50

0

500

-10

00

100

0-2

000

200

0-3

000

300

0-4

000

400

0-5

000

500

0-6

282

Nu

mb

er o

f T

TM

S S

ites

Number of Daily Truck Counts (Classes 8-13)

Freeways and Expressways

Divided Arterials

Undivided Arterials

Collectors

Centroid Connectors

One-Way Facilities

Ramps

Toll Facilities

using truck fleet data in combination with other …...using truck fleet data in combination with...

Documents